The ROI Revolution Blog

Using Website Optimizer with Google Analytics NEW!!

May 12, 2008


gwo-ga.gif
You may remember that back in April of '07, I came up with a way to get your Google Website Optimizer multivariate experiment data to show up in Google Analytics. While useful, there were a few drawbacks that I'm sure some of you have noticed, and it wasn't the easiest thing to implement. After getting a lot of great feedback from users, I've come up with a new script that has many advantages over the old method:

  • Uses easier implementation
  • Works with both ga.js and urchin.js (make sure you use the right instructions below for Step 2)
  • Includes error-handling so that a JavaScript error no longer occurs if an experiment is not yet running or is paused, stopped, or completed
  • Features automatic page name tracking - no more changing the Google Analytics code on the page
  • No longer replaces regular page reporting

As a refresher, the whole point of this integration is to allow you to make the most of your experiments. While Google Website Optimizer by itself can give you a quick look at which combination is best at improving conversion, it tells you nothing about transactions, revenue, micro-conversions, navigation, segmentation by source, and bounce rate. If you integrate Google Analytics into your Google Website Optimizer experiments, you will get much richer data, and be able to get a true idea of how your test is doing.

Again, this integration is designed for multivariate experiments only - you do not need to use any special tools to be able to get A/B test data from Google Analytics.

The first thing to do is find out if you are using ga.js or urchin.js. Depending on which version of the Google Analytics code you are using, you'll want to use different instructions.

Step 1: Upload the ga_gwo.js file to your server.

You can find this file here. Upload it to your server, and make a note of its location on your server, as you'll need it in the next step. That's it for Step 1!

Step 2: Paste the appropriate script onto your test page immediately after your existing Google Analytics code. There is one version of this for ga.js and another for urchin.js, so make sure you grab the right one!

If you are using ga.js, copy and paste the following code after your Google Analytics code:



You need to do three things to the script:

First, change the path to ga_gwo.js so that it points to the right location on your server.

Then, replace UA-XXXXXX-X with your Google Analytics account number.

Finally, change what's inside getcombo_ga to match your experiment. The numbers inside will match the number of sections and variations (including the original) that are in your experiment. So if you are running an experiment with 2 headlines, 3 images, and 4 paragraphs, again including the original, you would call getcombo_ga("2-3-4");

You can find a working example of a page that's using ga.js by viewing the source code of the page found here.

If you are using urchin.js, paste this code after your Google Analytics code instead:



You need to do two things to the script:

First, change the path to ga_gwo.js so that it points to the right location on your server.

Finally, change what's inside getcombo_urchin to match your experiment. The numbers inside will match the number of sections and variations (including the original) that are in your experiment. So if you are running an experiment with 2 headlines, 3 images, and 4 paragraphs, again including the original, you would call getcombo_urchin("2-3-4");

You can find a working example of a page that's using urchin.js by viewing the source code of the page found here.

That's it! Now you'll be tracking your combinations within Google Analytics!

Now, there are a few different ways you can go about seeing your data. I'd recommend excluding the test data from your main profile. You can do this by using a Custom Exclude Filter. Filter out any Request URIs that contain \?combo. You can simply create a duplicate profile that doesn't use this exclude filter.

Within your new profile, you would go to the Top Content report, where you can search for your test page by using the search bar on the bottom of the report. If you want to get more information about a specific combination, all you need to do is click on its name within the report.

Once you've clicked on a specific combination, you can then segment by a number of things. (As an aside, if you haven't seen Jeremy's cool Greasemonkey script for segmenting, you should definitely check it out!) Once you've segmented, you'll have access to the Site Usage, Goal Conversion, Ecommerce, and (for AdWords) Clicks tabs. There you can find all kinds of information about Bounce Rate, alternative goals, and Ecommerce data. You can then use these more detailed metrics to get the full story concerning your Google Website Optimizer experiment!

So feel free to use this tool and please leave your feedback so that I can continue making improvements. Thanks for reading, and best of luck with your testing!



If our experiences with paid search have taught us anything, it's that deliberate keyword lists, tightly structured accounts, and proven ads are only half of the equation. Intuitive landing pages tested with Google Website Optimizer experiments are just as essential for profitable PPC results. Learn more about how we maximize ROI from the search to the finalized sale.

Comments

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Luke Stevens said:

Awesome, thanks for posting this, the combined power of GWO and GA totally blows my mind. It's incredible to think how much power is there for free, all you need is a little elbow grease :)

May 14, 2008 8:57 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Luke:

I feel the same way. It's hard to believe that enterprise-level tools like these are available to absolutely everyone at no cost. That being said, you definitely have to roll up your sleeves sometimes, but it's absolutely worth it.

Thanks for reading, and please leave some feedback if you decide to use this!

May 15, 2008 8:27 AM

Vikas Gupta said:

In the previous article (old version) that referenced the old GA interface there was a way to see all the goal conversions by combination (http://www.roirevolution.com/blog/initialnavsmall.jpg). How do you see a similar report in the new GA interface?

As far as I see, top content only shows the $ Index instead of conversions and conversions by goal. Clicking through a URL on Top Content does not show goals (at least as far as I can tell either).

Thanks

May 20, 2008 5:42 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

You are were actually only one step away - when you click through a URL on Top Content, you then need to segment. You can segment by anything you want. Once you've segmented, you'll be able to see goal conversion and ecommerce metrics (you'll want to look at the totals - found just above the regular report data). I highly recommend adding one or more of these segmented reports to your dashboard, so you don't have to drill down every time.

We're developing some other possible ways to look at your combination data within Google Analytics, and I plan on blogging about landing page/normal page metrics within the next few days.

May 21, 2008 8:10 AM

Vikas Gupta said:

@Shawn
Thanks for the help! I found what I am looking for. I look forward to the upcoming info on other ways to look at combination data.

May 21, 2008 11:51 AM

Chris Goward said:

Great work as always, Shawn. Thanks for sharing this tool.

We love using your technology with our clients :)

Chris

May 21, 2008 3:26 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Chris:

Thanks for the compliment! I hope to come out with a few more things soon that I've been kicking around.

May 21, 2008 3:29 PM

Vikas Gupta said:

In the post you mentioned that you do not need to use this method to get AB test data out of GA. How does AB test data show up in GA?

Thanks
Vikas

May 30, 2008 5:53 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

A/B tests simply require you to have two different pages with two different names. If you have tagged both with Google Analytics code, then you can find metrics on A/B tests in the same way you find metrics for any other page on your site. You have slightly more options if the page you are testing is a landing page, but for any page on your site you can use the Top Content report and segmentation to find many important metrics. As a note - the Google Analytics code must come after the GWO control script at the top of your 'A' page to get best results.

June 2, 2008 7:51 AM

Vikas Gupta said:

Thanks again for the help!

June 3, 2008 2:15 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Vikas:

No Problem! Thanks for reading.

June 4, 2008 8:18 AM

Mike said:

I may have set it up incorrectly, but if I am varying just the headline with two variations on my homepage, I see three pages corresponding to this in GA:

/
/?combo=0
/?combo=1

Where the views of the last two add up to the views of the first.

Additionally, the bounce rate of the second two is 0%.

I think that GA sees the users as visiting "/" then moving to one of the combinations. Since this doesn't count as a single page visit, bounce rate isn't applicable, so the data we want is in "% Exit" instead.

Is this right? Have I set it up incorrectly.

Thanks alot for creating this free integration tool!

June 30, 2008 7:43 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Mike:

It sounds like you have it set up correctly. Currently, the Google Website Optimizer integration code fires a separate pageview, but we're looking into ways to make this better. For now, you can use the navigation summary and entrance paths report to find the data you're looking for, but I'll definitely look into making this easier for everyone.

That said, it should still function correctly for conversion rate, ecommerce, and other metrics. Bounce rate is the only one that should be affected by this.

Finally, to make sure you get Bounce Rate information for your original page ('/' in this case), you should exclude the two extra pageviews from your main Google Analytics profile. Thanks for the feedback and we hope to make this better in the future!

June 30, 2008 8:38 AM

Michael Altman said:

First, thanks for this post, it has been very helpful.

Second, we are running multiple optimizer tests at the same time in different areas of the site. I noticed that when multiple GWO tests are running that are integrated with GA, I get funky analytics data.

EXAMPLE
Lets say I'm running one test on mysite.com/page1 with 2 combinations and another test on mysite.com/page2 with 4 combinations. In GA I see analytics for mysite.com/page1?combo=3. This should not exsist since we are running only 2 combinations on this page.

Thanks for the help.

- Michael

August 11, 2008 10:12 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Michael:

Right now the Google Website Optimizer integration script is only designed to work with one experiment running at a time. It's definitely possible to rewrite the code of the tool to account for more than one experiment, and that's something I will look into when the tool gets updated again. Unfortunately, with my current workload, I can't promise that it will happen soon.

If you're good with JavaScript, you may be able to manipulate the code yourself to account for this, although you'd have to also know a lot about how Google Website Optimizer cookies are formatted, which can be tricky. Alternatively, I plan on updating this tool sometime in the future, although unfortunately I can't give you a solid date for that.

August 11, 2008 12:24 PM

Scott said:

** A/B TESTING **

I know this isn't for use with A/B tests, but...

Oddly, when I run an A/B test and track stats in Google Analytics, I'm getting some funky data as well, such as:

100% Bounce rates on two out of the four landing pages being tested, with only around 50% exit rate.

Bounce rate is significantly low on the original page. Same for Time on Page.

Original page also gets 10x - 100x more "visits" within "Top Landing Pages" than other pages.

** PROBABLY CAUSE **

This data tends to suggest one thing:

The method Website Optimizer uses to "randomize" A/B pages (by forwarding to the final landing page from the original landing page) is causing the stats to get logged for both the original page hit and the final landing page hit, after being forwarded to it by Website Optimizer.

I wonder if I missed something when setting up tracking for both WO and GA. :/

--
Note: these are all new pages, created solely for use by the test, and linked only from our CPC campaigns. We aren't running more than one test either.

September 12, 2008 12:07 PM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Scott:

Sounds like there's a problem with the order of your Google Analytics and Google Website Optimizer codes on your original and test pages. Bounce Rate and Visits should remain unaffected by an A/B test, unless there is an error with implementation.

The Google Website Optimizer code that chooses which page shows up does not log anything within Google Analytics - it's completely separate. The first thing I would do is make sure that the code is in the following order on your original page:

Google Website Optimizer Control Script
Google Analytics Script
Google Website Optimizer Tracking Script

And on the test pages:

Google Analytics Script
Google Website Optimizer Tracking Script

Also, make sure you are using the correct version of the Google Website Optimizer script. If you are using ga.js for Google Analytics, you should also use ga.js for Google Website Optimizer. There are also some additional rules you need to follow when using both Google Analytics and Google Website Optimizer at the same time that may apply to you.

Other than that - I can't tell you exactly what's going on without seeing the pages myself. There may be something else that is causing problems with your setup, but I'd start with the things mentioned above. Hope that helps!

September 12, 2008 1:10 PM

Ophir said:

If you're using this hack to help you measure multiple conversion events for a single experiment (using GA goals) you can check the conversion rate confidence level using this online calculator.

http://www.prusak.com/tools/conversion-confidence-calculator/

October 20, 2008 2:02 PM

Tomas Vaitulevicius said:

Hi Guys, Thanks for a great script! We're working to implement it right now...we still have some .js compatibility issues, but we'll get that sorted.
The real problem that we have is the fact that our site gets more than half a million visits and more than 15 mill pageviews every day. Because of that all segmented (dimensions) data that you get for any top content item on GA is sampled and Totally Wrong. That means that we have no clue how many test goal-conversions did the variations have or how many real-site conversions did the visitors that got different variations do.
As you are already thinking about moving away from using trackpageview for this script, I wonder if you consider using SetVars (User Defined) or Event tracking for this that?
Cheers, T

January 30, 2009 5:23 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Tomas:

While we're certainly investigating any and all possibilities, the _trackPageview method has worked very well for us.

While using the dimension dropdown will give you sampled data if you have a massive amount of traffic, there are a few things you can try:

1. Try to look at as small a sample of data as possible - maybe like 1 day at a time? With 500,000 visitors a day, this still might not work.
2. Use the new 'Advanced Segments' feature within Google Analytics. You can create an advanced filter to look at data for all traffic that landed on a specific page, and that data may be a little more accurate than the sampled data that you are getting with the dimensions dropdown.

You can also use the Site Search-compatible code that Michael outlines in an even more recent version of this method, which can be found here. Then you can set up a profile that uses 'combination' as it's site search parameter to get conversion numbers on a session by session basis.

January 30, 2009 8:49 AM

Dan Shields said:

Just wondering, I know you say you can only use it for one experiment at a time for a given site. Does this mean if I have an account that I use for 20 different sites and one of those sites don't have any experiments running I still can't use it because I have other experiments running for other sites with the same account or is it site specific?

February 4, 2009 11:54 AM

Shawn Purtell, Senior Web Analytics Engineer Author Profile Page said:

@Dan:

The code has actually been updated. You can find the newest version here. If you use the new code, you can actually get multivariate data for multiple experiments within Google Analytics. This should help out.

February 5, 2009 10:50 AM

Nosy Rat said:

Will give it a try. Thanks.

July 1, 2010 6:37 AM

Post Your Comments

© 2002-2014 ROI Revolution, Inc. All rights reserved.