How To Use Google Analytics New Attribution Modeling Tool (And how to get it)

by 16 02/07/2013

One of the knocks on Google Analytics has been the way it handles goal attribution.

Translation:  the way that Google Analytics currently gives credit to a conversion (a purchase, a form fill, a donation, etc) can be very misleading.

Google Analytics currently attributes credit for a conversion using the Last Interaction Model.  If you want the long story on this read this article on Attribution Models.

Here’s an example of the Last Interaction Attribution Model at work…

As of this writing, Google Analytics Attribution Modeling tool is not available to everyone.  You can get access by putting your name on this list.  They won’t email you when they add it to your reports, it will just appear.  So, check back once in a while.  It took a couple of weeks for them to add it to my reports.
  1. Jane is researching blue buckets and finds your website via Google Search.  She browses around and likes the buckets you sell.  But she’s not ready to buy.  She clicks to “Like” your company on Facebook.  She then leaves your website.
  2. Your company writes an article on your company blog about how one of your customers is using your blue buckets successfully.  You post a link to the article on your Facebook page.  Jane clicks on this link and reads the article.  She finds the article very interesting.  She exits your website.
  3. Jane is back on Google researching blue buckets and sees a PPC ad to your website.  She clicks on it.  She buys a blue bucket.

Attribution for this scenario would look like this in the Last Interaction model,

Last Interaction Model

This causes obvious problems when trying to decide where, when and how much to spend on a marketing channel or tactic.  In the above example, we might make the decision to cut the SEO and social media budgets and increase the PPC budget.

That is, until now.

New Attribution Modeling Tool provides more clarity

The Last Interaction Model is still available in the Google Analytics Attribution Modeling Tool but now there are a number of other models to choose from as well.

Here is a look at the new attribution models available in Google Analytics,

attribution-models

If, in our blue buckets example above you chose to look at your reports using the First Interaction Model, it would look like this,

First Interaction Attribution Model

 

In other words, the First Interaction would get all the credit.  You might use this to find out what channels are driving awareness (top of funnel) activity for you.   I’ll show an example of that in just a minute.

For more information about different types of attribution models, read this.

Google Analytics Attribution Modeling Tool In Action

As with any blog, our lifeline is our audience.  Growing subscribers is a high priority for us — but what are the primary drivers for driving subscribers for The Daily Egg?

The new attribution modeling tool in Google Analytics gives us several ways to think about the way our marketing is driving subscribers,

interaction-models

The Last and First Interaction Models are fairly straight forward.  Time Decay is an attribution model that gives more attribution to interactions that happened recently and less to those that happened further in the past.

As you can see with the above report, Referral Traffic is a popular way for those that eventually subscribe to the blog to become aware of the blog.

Let’s say we want to take a look at how well some of our referral sources  are doing at driving Free Trial sign-ups for the Crazy Egg software.

Let’s take a look at how well QuickSprout (Neil Patel’s blog) and The Daily Egg (this blog) are doing at driving sign-ups.

referrals-from-blog

As you can see, these blogs are better at introducing folks to a product than they do closing the deal.  There is about a 90% increase in conversions for QuickSprout when you give 100% attribution to the first interaction — something that would be completely lost in Google Analytics before the new Attribution Modeling Tool.

It’s not perfect (far from it)

It seems I end a lot of analytics articles the same way — it’s not perfect.  Analytics aren’t perfect and they never will be.

This report, like any other GA report, can be skewed by people deleting cookies, moving devices, blocking javascript, etc.

But one thing that keeps this tool from being far from perfect is something called the Lookback Window.

lookback-window

 

The tool is only taking into account interactions that took place in the last 30 days.

For example, if you were looking at conversions using a First Interaction model it might look like this,

look-back-window-problem

 

Since the tool doesn’t look back more than 30 days, your SEO gets no credit for this conversion — even in a First Interaction Model.  In this case, the First Interaction would be counted as the Social Network.

I’ve heard rumors that Google Analytics is going to increase the window to 90 days at some point, but for now we are working with 30 days of data.

But, as with all things in Analytics, we are looking at the big picture.  And these new methods for viewing goal attributions are, all in all, a great and welcome addition to the free version of Google Analytics.

Print Friendly

About 

Russ Henneberry is the Editorial Director at Digital Marketer. He's worked on digital marketing projects for companies like CrazyEgg, Salesforce.com and Network Solutions. You can connect with Russ on Twitter, LinkedIn, Google+ or on his blog.

Get our Daily Newsletter

Get conversion optimization, design and copywriting articles delivered to your inbox FREE

16 COMMENTS

Krishnan

Hey Russ,

Thanks for the link to the whitelist application. Weren’t the models only available to premium members?

February 7, 2013 Reply

    Russ Henneberry

    Hi Krishnan. Correct, Google Analytics Premium has had these models for some time now. However, if you don’t have 150K per year for GA Premium, the whitelist is a good way to go. :)

    February 7, 2013 Reply

Lance Cummins

Thanks for the helpful article on this. I’m hoping I can get access to the tool and begin to get a clearer picture of attribution for conversions. Keep up the great writing!

February 7, 2013 Reply

Vernon Riley

Great article – and the ability to define custom conversion segments where you can define how you want the attributions to be filtered and displayed is really good.

February 8, 2013 Reply

    Russ Henneberry

    Thanks Vernon — I have yet to play with that functionality — but I plan to very soon. Thanks for taking the time to mention this!

    February 8, 2013 Reply

Charles Davis

Nice job introducing the Attribution Modeling tool. Everyone has access to the Multi-channel Funnel reports and Attribution Modeling makes getting insights out of it much more powerful. I recently posted a detailed article about the tool that includes how to use the custom channel groupings and custom model builder: http://www.blastam.com/blog/index.php/2013/01/what-is-google-attribution-modeling-do-you-need-it/

The ability to customize this tool makes it pretty powerful. At the same time, you’re correct, it’s not perfect and never will be. But for a vast majority of people/companies that don’t have access to any attribution modeling tool currently this is a big upgrade.

Also, the extension to a 90 day look back window is not a rumor. Google confirmed it at the Google Analytics Partner Summit last year. Who knows when it will happen though. Like you said, it’ll just be there one day. Just keep checkig for it.

February 8, 2013 Reply

    Russ Henneberry

    Charles, thanks so much for dropping by and giving us your take on the tool. Also, that article is absolutely outstanding — it’s one of the articles I read in preparation for this article. Well done.

    February 11, 2013 Reply

Ken P.

From a media mix perspective it works great. The only issue I have right now with this tool is the inability to view interactions on the transaction level. The developer API also has no reference to pulling this data. One could asume if you have reached the point of measuring and modelling the customer level journey your needs have surpassed the platform.

March 27, 2013 Reply

Juanjo

Hi Russ,
I want to use this reports but I have founded some differences when I compare traffic sources report (e-commerce) with reports of transaction with “Attribution modeling tool” using last interaction. Do you know why this happen?
Best!
JJ

May 10, 2013 Reply

    Tyson

    Hi JJ,

    Thought I might jump in and help out a little…it can definitely be frustrating to compare the Last Interaction reports in the AMT with the standard GA reports – they will alwyas be close but never match exactly! The reason is because standard GA reports don’t use a strict last interaction model.

    The standard GA attribution model is technically “last-non-direct-interaction” with a 6-month lookback window. So if a visitor comes in through email and then comes back two days later directly and purchases, the standard GA reports will credit the email channel while the AMT tool will credit Direct when using the Last Interaction model.

    This is definitely frustrating…hopefully Google gives us the ability to choose a global attribution model in the future that would apply uniformly to all reports. This is definitely one of the things that confuses our clients on a regular basis.

    Hope that helps…

    May 29, 2013 Reply

Ecommspark

Thank you very much Russ for this article. Article is very educational and easy to read. The topic of the article the attribution models is the most hot topic right now in online industry since it’s kind of new, still evolves and not everyone know what it is, how it’s being applied and what you can learn from it. Many companies don’t even know from where to start and don’t understand the importance of the attribution models.

September 27, 2013 Reply

Steve

It’s reassuring to see these gaps in analytics don’t just affect us, especially ‘direct’ which is somewhat of a black hole to us and many others.

However, looking at the last click attribution model in GA’s model comparison section; shouldn’t this number tally up perfectly with the value you see in goals for a particular channel/source/medium, as the goal attribution works on last click?

The reason I ask is because the 2 values in analytics are different for us and we’re trying to get to the bottom of just how many conversions our PPC is truly responsibile for (top of the funnel). If goals work on last click, and we observe last click in the model comparison – why would the values differ?

April 9, 2014 Reply

    Tyson

    Hey Steve,

    This is a good (and common) question. The key is, the default model in Google Analytics is not completely last click. It’s actually last non-direct click. Also, the conversion windows are going to differ from the reports in MCF and the standard reports. It would be nice if Google Analytics standardized this a little better, but that’s two reasons for the discrepancies you are seeing.

    April 9, 2014 Reply


Leave comment

Some HTML allowed

Get conversion optimization & A/B testing articles FREE >>>