The ROI Revolution Blog

Understanding Correlations in Google Analytics

May 28, 2009

Website traffic does not exist in a vacuum. Here's a simple example: Jack comes to your website on Monday after seeing one of your AdWords content ads and he likes what he sees. He's a careful shopper though, so he's not ready to commit quite yet and leaves the site for the day. He takes some time and does some comparison shopping throughout the week, talks to some of his friends and comes back to your site again after typing your company name into Google and clicking on an organic result. He sees an offer for a 10% off coupon if he signs up for your newsletter, so he does, and then leaves the site again. In a week, he gets an email about a sale you are having, and clicks on a link within the email, finally making a purchase on this, his third visit.

sp-corr-title.jpg

So the big question is - how does this show up in Google Analytics? Does AdWords get any credit for the sale? The simple answer is no. Depending on if you are tracking your emails in Google Analytics (and how you are doing it), you'll either see a conversion for the email, the organic branded search or even a direct visit. Wouldn't it be nice to know that at one point AdWords had something to do with the sale? Better yet, wouldn't you like to know the Campaign, Ad Group and Keyword that was responsible?

Jack's example is a very common one, and pretty simple in comparison to the way a lot of people use the Internet, so it is important to try and understand the relationships between your different traffic sources.

Still don't care? Let me give you a real-world example of what can happen if you ignore it:

Case Study: A company that deals in a software product noticed that it was getting what looked like a pretty poor return on Content Network traffic from AdWords (responsible for what Google Analytics reported as roughly 5% of daily revenue). In an attempt to reduce costs, they decided to pause this traffic completely. The result was that almost immediately they noticed a 15-20% drop in daily revenue!

What the heck happened? Well, it turns out a large percentage of that content traffic was coming back as either organic branded traffic or direct traffic. They never bothered to look at the relationship between their content traffic and other traffic sources, and it cost them.

Conversely, by understanding this relationship, they have been able to not only gain back the 15-20% that they lost, but improve the return even further!

So how can you learn from their mistake? Here are a few things you can do, ranging from fairly simple to more complex, to help you grasp the relationship between your marketing sources and mediums:

Method #1: Look for basic correlations.

Depending on your site and where you are spending your advertising dollars, your traffic will have a different dynamic. Since my specialty is in dealing with Paid Search traffic from Google AdWords, Yahoo Search Marketing, and MSN AdCenter, I am going to focus on analyzing results from AdWords.

The two most common sources that will see a bump from AdWords traffic are direct visitors and organic traffic from branded keywords (and possibly email visitors).

There are a few ways that you can look and see if there is any kind of relationship between these different kinds of traffic. The first and most basic is to use the Google Analytics timeline to compare date ranges for these different kinds of traffic.

First, select a date range you would like to compare. In this case, we're doing a week-to-week comparison:

sp-daterange-comp.png

Second, you can choose to show the Medium from the dropdown list. This is optional, but since we're looking at really top-level data, it does the job:

sp-medium-dropdown.png

Here's what we saw in the report:

sp-report1.png

Now, we're simply looking for clues here - nothing set in stone. We just want to know if there's a possibility that cpc traffic is affecting direct and organic traffic. We'll do more digging in a bit, but as you can see from the trend, it looks like all three sources trended down from the previous week. This simply means that there is an opportunity to learn more. This is something you most likely want to check again if you make any major changes to your AdWords strategy.

If you have the time, you can do this at a more granular level, by breaking it down into day-by-day trends. You'd get something like this:

sp-trend-daily-data.png

sp-trend-daily-graph.png

Looks like there is some kind of relationship there, doesn't it?

You can also see basic trends by using Advanced Segments feature of Google Analytics. You can select the following segments:

sp-adv-segments.png

to see the data over time for all three mediums.

sp-adv-seg-graph.png

Note that if your traffic is unbalanced (i.e. 80% cpc), this graph may be a little harder to read, in which case I recommend exporting the data into Excel to get results similar to the ones I've posted above.

Either way, it still seems clear that there is some kind of relationship here. This of course begs the question, what is that relationship and how can I quantify it? For that, you have some more options.

Method #2: Use the User Defined Variable within Google Analytics.

This method will provide you will a much more specific level of analysis than method #1, although the two can be used in conjunction. Typically, you would do this after checking to see if there are any correlations by doing a more general analysis. I'm going to cover this strategy as it pertains to AdWords, although you could certainly modify this strategy to do even more robust analysis.

The idea is that whenever a visitor comes to your site via AdWords (or YSM or MSN), you set the User Defined variable for Google Analytics that does not get overwritten. You obviously have to do some coding to accomplish this, but it's fairly straightforward.

Even if the regular Google Analytics cookies are overwritten, you still have the User Defined variables hanging around unchanged so that you can see which percentage of direct and organic (and other) traffic has at one point clicked on one of your ads.

sp-segment-ud.png

You do this by using segmentation within your All Traffic Sources report. First, I recommend changing the view to Medium again, and then clicking on one of the non-cpc mediums on the list. For example, if I segment direct traffic by User Defined, I might see the following report:

sp-ud-direct-report.png

Because this report only represents direct traffic, and that 'not set' represents people who have never been tagged as having come from cpc, that means that a whopping 41% of all direct traffic for this time period has at one point come from AdWords! Not only that, but if we wanted to go over to the Ecommerce tab, we would get a feel for whether or not those visitors are more valuable than normal direct traffic.

Eventually, you'll be able to break down all of your traffic in this way:

sp-percent-report.png

What you are seeing is the residual effects of AdWords traffic - traffic you otherwise would have missed entirely. If you want to take it a step further, you could include the Campaign, Ad Group, and Keyword in your User Defined variable and get even more bang for your buck!

Even though this is a great way to get more data, there still are some other questions. What about people who use multiple computers? Or delete cookies? What can you do? I won't go into too much detail, but the idea is similar.

Method #3: Use a custom system.

This really depends on what's possible with your internal customer organization systems, but there are a variety of things you can do here. One I've seen is to keep a custom cookie for users that is then associated with them when they create an account for your store (or sign up for your newsletter). This way, even if they come back from a different computer later, you can see their entire history and associate it with a keyword.

I've seen a lot of different variations of this system - but the vast majority of this information will still be limited by the same things that limit web analytics in general - mainly cookie blocking/deletion and JavaScript disabling.

That's it! Feel free to customize these ideas to fit your own needs.

Google Analytics for Online Advertisers
Here at ROI Revolution, we consider Google Analytics tracking essential for paid search, so it's included in our PPC Campaign Management service.

TrackBack

Listed below are links to weblogs that reference Understanding Correlations in Google Analytics:

» Blogzwonull Links der Woche: Lesenswertes der KW21 bis KW24/2009 from BlogZwoNull
Da ich nun nach etwas längerer Pause mal wieder dazu gekommen bin meinen Feedreader mit den gemerkten Artikeln der letzten Wochen auszuleeren, gibt es heute wieder die Blogzwonull – Links der Woche: Conversion Optmization: How to optimize y... [Read More]

Tracked on June 15, 2009 1:09 AM

Comments

adriano nadalin said:

hi!

I would first like to say that a realy liked the post.

I once used the setVar function to segment a trafic source, i used to use it on the first page view of the visitor so i could set a user defined variable of the visitor from an especific trafic source.

That had a big impact on bounce rate of my site, and after some research i understood that this happens because i was using the setVar on the first page viewed by the visitor.

Could this be a problem to set a user defined variable when the visitor comes from adwords?

The "Method #2: Use the User Defined Variable within Google Analytics" would be a good way to see this correlation of traffic sources but i am afraid it would impact other data as bounce rate of my site.

Tks!

May 29, 2009 3:50 PM

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

@Adriano:

I'm happy to report that this problem has been fixed for a little while now. Setting the User Defined variable no longer affects Bounce Rate!

May 29, 2009 5:38 PM

Beate Vervaecke said:

I've been thinking about implementing method 2.
But as far as I understand the given example above, in combination with the cookie timing and "priority", using this method for assessing the real source of direct traffic is only reliable if your CPC-campaign has ended more then 6 months ago?
Otherwise the __utmz would still state CPC or another source?
What do I miss?

June 1, 2009 6:30 AM

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

@Beate:

Well, there are many reasons why a returning visitor would show up as direct even if they had been to the site less than 6 months ago. The primary cause of this is an error with Google Analytics code on the site or an improperly linked Google Analytics/AdWords account, but there are several other reasons why this would happen.

The point is, you may not see (and probably won't see) numbers as high as 40% for direct traffic, but if you have a lot of cpc traffic, you'll still see some. Usually, I see this somewhere in the neighborhood of 5 to 15%.

You already know about the visitors tagged as 'cpc' - what we really care about here is the traffic that is less obvious.

For non-direct traffic sources, it's even more important, because those will be overwritten because of how Google Analytics works with returning visitors. This includes organic traffic, email traffic, banner traffic, referring traffic, and even in some cases other CPC traffic.

June 1, 2009 7:58 AM

Allaedin Ezzedin said:

Thanks Shawn... great post!

As for cookies blocking/deletion, analysts should always assume visitors do not clear their cookies and even if they do (indeed many do) who cares since we are usually after analyzing trends and not exact numbers.

June 1, 2009 4:08 PM

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

@Allaedin:

I agree, but it still is something to keep in mind, especially when looking at any data that depends on the use of cookies. Thanks for reading!

June 1, 2009 4:39 PM

adriano nadalin said:

hi!

I would first like to say that a realy liked the post.

I once used the setVar function to segment a trafic source, i used to use it on the first page view of the visitor so i could set a user defined variable of the visitor from an especific trafic source.

That had a big impact on bounce rate of my site, and after some research i understood that this happens because i was using the setVar on the first page viewed by the visitor.

Could this be a problem to set a user defined variable when the visitor comes from adwords?

The "Method #2: Use the User Defined Variable within Google Analytics" would be a good way to see this correlation of traffic sources but i am afraid it would impact other data as bounce rate of my site.

Tks!

May 29, 2009 3:50 PM

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

@Adriano:

I'm happy to report that this problem has been fixed for a little while now. Setting the User Defined variable no longer affects Bounce Rate!

May 29, 2009 5:38 PM

Beate Vervaecke said:

I've been thinking about implementing method 2.
But as far as I understand the given example above, in combination with the cookie timing and "priority", using this method for assessing the real source of direct traffic is only reliable if your CPC-campaign has ended more then 6 months ago?
Otherwise the __utmz would still state CPC or another source?
What do I miss?

June 1, 2009 6:30 AM

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

@Beate:

Well, there are many reasons why a returning visitor would show up as direct even if they had been to the site less than 6 months ago. The primary cause of this is an error with Google Analytics code on the site or an improperly linked Google Analytics/AdWords account, but there are several other reasons why this would happen.

The point is, you may not see (and probably won't see) numbers as high as 40% for direct traffic, but if you have a lot of cpc traffic, you'll still see some. Usually, I see this somewhere in the neighborhood of 5 to 15%.

You already know about the visitors tagged as 'cpc' - what we really care about here is the traffic that is less obvious.

For non-direct traffic sources, it's even more important, because those will be overwritten because of how Google Analytics works with returning visitors. This includes organic traffic, email traffic, banner traffic, referring traffic, and even in some cases other CPC traffic.

June 1, 2009 7:58 AM

Allaedin Ezzedin said:

Thanks Shawn... great post!

As for cookies blocking/deletion, analysts should always assume visitors do not clear their cookies and even if they do (indeed many do) who cares since we are usually after analyzing trends and not exact numbers.

June 1, 2009 4:08 PM

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

@Allaedin:

I agree, but it still is something to keep in mind, especially when looking at any data that depends on the use of cookies. Thanks for reading!

June 1, 2009 4:39 PM

adriano nadalin said:

hi!

I would first like to say that a realy liked the post.

I once used the setVar function to segment a trafic source, i used to use it on the first page view of the visitor so i could set a user defined variable of the visitor from an especific trafic source.

That had a big impact on bounce rate of my site, and after some research i understood that this happens because i was using the setVar on the first page viewed by the visitor.

Could this be a problem to set a user defined variable when the visitor comes from adwords?

The "Method #2: Use the User Defined Variable within Google Analytics" would be a good way to see this correlation of traffic sources but i am afraid it would impact other data as bounce rate of my site.

Tks!

May 29, 2009 3:50 PM

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

@Adriano:

I'm happy to report that this problem has been fixed for a little while now. Setting the User Defined variable no longer affects Bounce Rate!

May 29, 2009 5:38 PM

Beate Vervaecke said:

I've been thinking about implementing method 2.
But as far as I understand the given example above, in combination with the cookie timing and "priority", using this method for assessing the real source of direct traffic is only reliable if your CPC-campaign has ended more then 6 months ago?
Otherwise the __utmz would still state CPC or another source?
What do I miss?

June 1, 2009 6:30 AM

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

@Beate:

Well, there are many reasons why a returning visitor would show up as direct even if they had been to the site less than 6 months ago. The primary cause of this is an error with Google Analytics code on the site or an improperly linked Google Analytics/AdWords account, but there are several other reasons why this would happen.

The point is, you may not see (and probably won't see) numbers as high as 40% for direct traffic, but if you have a lot of cpc traffic, you'll still see some. Usually, I see this somewhere in the neighborhood of 5 to 15%.

You already know about the visitors tagged as 'cpc' - what we really care about here is the traffic that is less obvious.

For non-direct traffic sources, it's even more important, because those will be overwritten because of how Google Analytics works with returning visitors. This includes organic traffic, email traffic, banner traffic, referring traffic, and even in some cases other CPC traffic.

June 1, 2009 7:58 AM

Allaedin Ezzedin said:

Thanks Shawn... great post!

As for cookies blocking/deletion, analysts should always assume visitors do not clear their cookies and even if they do (indeed many do) who cares since we are usually after analyzing trends and not exact numbers.

June 1, 2009 4:08 PM

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

@Allaedin:

I agree, but it still is something to keep in mind, especially when looking at any data that depends on the use of cookies. Thanks for reading!

June 1, 2009 4:39 PM

adriano nadalin said:

hi!

I would first like to say that a realy liked the post.

I once used the setVar function to segment a trafic source, i used to use it on the first page view of the visitor so i could set a user defined variable of the visitor from an especific trafic source.

That had a big impact on bounce rate of my site, and after some research i understood that this happens because i was using the setVar on the first page viewed by the visitor.

Could this be a problem to set a user defined variable when the visitor comes from adwords?

The "Method #2: Use the User Defined Variable within Google Analytics" would be a good way to see this correlation of traffic sources but i am afraid it would impact other data as bounce rate of my site.

Tks!

May 29, 2009 3:50 PM

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

@Adriano:

I'm happy to report that this problem has been fixed for a little while now. Setting the User Defined variable no longer affects Bounce Rate!

May 29, 2009 5:38 PM

Beate Vervaecke said:

I've been thinking about implementing method 2.
But as far as I understand the given example above, in combination with the cookie timing and "priority", using this method for assessing the real source of direct traffic is only reliable if your CPC-campaign has ended more then 6 months ago?
Otherwise the __utmz would still state CPC or another source?
What do I miss?

June 1, 2009 6:30 AM

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

@Beate:

Well, there are many reasons why a returning visitor would show up as direct even if they had been to the site less than 6 months ago. The primary cause of this is an error with Google Analytics code on the site or an improperly linked Google Analytics/AdWords account, but there are several other reasons why this would happen.

The point is, you may not see (and probably won't see) numbers as high as 40% for direct traffic, but if you have a lot of cpc traffic, you'll still see some. Usually, I see this somewhere in the neighborhood of 5 to 15%.

You already know about the visitors tagged as 'cpc' - what we really care about here is the traffic that is less obvious.

For non-direct traffic sources, it's even more important, because those will be overwritten because of how Google Analytics works with returning visitors. This includes organic traffic, email traffic, banner traffic, referring traffic, and even in some cases other CPC traffic.

June 1, 2009 7:58 AM

Allaedin Ezzedin said:

Thanks Shawn... great post!

As for cookies blocking/deletion, analysts should always assume visitors do not clear their cookies and even if they do (indeed many do) who cares since we are usually after analyzing trends and not exact numbers.

June 1, 2009 4:08 PM

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

@Allaedin:

I agree, but it still is something to keep in mind, especially when looking at any data that depends on the use of cookies. Thanks for reading!

June 1, 2009 4:39 PM

adriano nadalin said:

hi!

I would first like to say that a realy liked the post.

I once used the setVar function to segment a trafic source, i used to use it on the first page view of the visitor so i could set a user defined variable of the visitor from an especific trafic source.

That had a big impact on bounce rate of my site, and after some research i understood that this happens because i was using the setVar on the first page viewed by the visitor.

Could this be a problem to set a user defined variable when the visitor comes from adwords?

The "Method #2: Use the User Defined Variable within Google Analytics" would be a good way to see this correlation of traffic sources but i am afraid it would impact other data as bounce rate of my site.

Tks!

May 29, 2009 3:50 PM

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

@Adriano:

I'm happy to report that this problem has been fixed for a little while now. Setting the User Defined variable no longer affects Bounce Rate!

May 29, 2009 5:38 PM

Beate Vervaecke said:

I've been thinking about implementing method 2.
But as far as I understand the given example above, in combination with the cookie timing and "priority", using this method for assessing the real source of direct traffic is only reliable if your CPC-campaign has ended more then 6 months ago?
Otherwise the __utmz would still state CPC or another source?
What do I miss?

June 1, 2009 6:30 AM

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

@Beate:

Well, there are many reasons why a returning visitor would show up as direct even if they had been to the site less than 6 months ago. The primary cause of this is an error with Google Analytics code on the site or an improperly linked Google Analytics/AdWords account, but there are several other reasons why this would happen.

The point is, you may not see (and probably won't see) numbers as high as 40% for direct traffic, but if you have a lot of cpc traffic, you'll still see some. Usually, I see this somewhere in the neighborhood of 5 to 15%.

You already know about the visitors tagged as 'cpc' - what we really care about here is the traffic that is less obvious.

For non-direct traffic sources, it's even more important, because those will be overwritten because of how Google Analytics works with returning visitors. This includes organic traffic, email traffic, banner traffic, referring traffic, and even in some cases other CPC traffic.

June 1, 2009 7:58 AM

Allaedin Ezzedin said:

Thanks Shawn... great post!

As for cookies blocking/deletion, analysts should always assume visitors do not clear their cookies and even if they do (indeed many do) who cares since we are usually after analyzing trends and not exact numbers.

June 1, 2009 4:08 PM

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

@Allaedin:

I agree, but it still is something to keep in mind, especially when looking at any data that depends on the use of cookies. Thanks for reading!

June 1, 2009 4:39 PM

adriano nadalin said:

hi!

I would first like to say that a realy liked the post.

I once used the setVar function to segment a trafic source, i used to use it on the first page view of the visitor so i could set a user defined variable of the visitor from an especific trafic source.

That had a big impact on bounce rate of my site, and after some research i understood that this happens because i was using the setVar on the first page viewed by the visitor.

Could this be a problem to set a user defined variable when the visitor comes from adwords?

The "Method #2: Use the User Defined Variable within Google Analytics" would be a good way to see this correlation of traffic sources but i am afraid it would impact other data as bounce rate of my site.

Tks!

May 29, 2009 3:50 PM

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

@Adriano:

I'm happy to report that this problem has been fixed for a little while now. Setting the User Defined variable no longer affects Bounce Rate!

May 29, 2009 5:38 PM

Beate Vervaecke said:

I've been thinking about implementing method 2.
But as far as I understand the given example above, in combination with the cookie timing and "priority", using this method for assessing the real source of direct traffic is only reliable if your CPC-campaign has ended more then 6 months ago?
Otherwise the __utmz would still state CPC or another source?
What do I miss?

June 1, 2009 6:30 AM

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

@Beate:

Well, there are many reasons why a returning visitor would show up as direct even if they had been to the site less than 6 months ago. The primary cause of this is an error with Google Analytics code on the site or an improperly linked Google Analytics/AdWords account, but there are several other reasons why this would happen.

The point is, you may not see (and probably won't see) numbers as high as 40% for direct traffic, but if you have a lot of cpc traffic, you'll still see some. Usually, I see this somewhere in the neighborhood of 5 to 15%.

You already know about the visitors tagged as 'cpc' - what we really care about here is the traffic that is less obvious.

For non-direct traffic sources, it's even more important, because those will be overwritten because of how Google Analytics works with returning visitors. This includes organic traffic, email traffic, banner traffic, referring traffic, and even in some cases other CPC traffic.

June 1, 2009 7:58 AM

Allaedin Ezzedin said:

Thanks Shawn... great post!

As for cookies blocking/deletion, analysts should always assume visitors do not clear their cookies and even if they do (indeed many do) who cares since we are usually after analyzing trends and not exact numbers.

June 1, 2009 4:08 PM

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

@Allaedin:

I agree, but it still is something to keep in mind, especially when looking at any data that depends on the use of cookies. Thanks for reading!

June 1, 2009 4:39 PM

Post Your Comments

© 2002-2014 ROI Revolution, Inc. All rights reserved.