Web Metrics | Search Marketing
Site Strategy

"In a nutshell, Jason created our marketing analytics capability. He was able to figure out what data we collect, where it is, what was missing, and hook it all up so we canget meaningful, actionable data. Our marketing efforts have improved leads and conversions in some cases by an order of magnitude. He knows his stuff."
Chris Foleen, Marketing Project Coordinator, TransCore, Inc.


Thursday, April 15, 2010

Which Touch? Going Past Last Touch Attribution in Google Analytics, a response

The guys at SEOMOZ have done some great thinking in their article "How to get past last touch attribution with Google Analytics" describing how to tinker with Google Analytics to get more than last touch attribution out of organic search referrals.  

Referencing Avinash's post on the value of "Upper funnel" keywords and several very interesting articles on how to build custom filters, they lay out a theory for how this could work.

"First touch attribution" is the first topic they take on.  Here the theory is that first touch keywords (upper funnel keywords) are the real stars.  Since they do the introducing they should be getting conversion credit.  Repeat visitors who use branded phrases and convert are only there because of the topic first touch.  This accounts for the higher conversion rate of branded search repeat visitors, but credit should really be given to the first touch.  The SEOMOZ guys do an interesting job of showing how to create the filters and settings to show first touch attribution in Google Analyics.

This is an interesting proposition and worth investigating.  However, I am skeptical.  In my experience, low priced products convert better from topical search while high priced products convert better from branded search.  For topical search conversions, we don't need to dig any further.  But branded higher priced purchases speak to the power and improtance of brand in the psychology of the buying cycle. This is ignored (maybe even discounted) by first touch attribution and the SEOMOZ article.  Even if topical search is attracting new visitors, brand is an undenighable influencer in big ticket and/or multi-touch purchasing decisions.  If you were to start taking your eye off your branded, repeat traffic in favor of first touch metrics, you could very well misunderstand your audience and their motivations.

The only way to work out the value of this and how to weight the metrics would be to run a test and see the results.  The test would have to collect all touch data so it could be evaluated.  This gets to the second part of the SEOMOZ article--multi-touch attribution.  

"Multi-touch attribution" is involved collecting all the referral data and then, normally, weighting it. First touch and last touch are the main points of interest.  The content and quality of the touches in between are informative.  Everything is given an "assist" value to weight its importance.

Here the SEOMOZ article has an interesting, but most-likely unworkable solution, especially if your website has any real volume of traffic.  

I appauld them for pointing out how it is possible by using a combination of filters and scripting to collect this information.  I also applaud them for recognizing that this would absolutely require a data export and external manipulation.   However, where they say a pivot table, I'm betting it would be more like a database and some creative SQL if your website has any volume at all.  This is because by concatenating the sources together into the _setCustomVar value in Google Analytics, you are essentially creating automatic segmentation and your variations are going to multiply uncontrollably.  Eventually you could have a quantity of segments equal to the number of visitors you have.  Only those who followed exactly the same series of keywords and sources could be aggregated together.  Because of this, I have a hard time seeing how this could be applied in any way outside a test environment unless more software and infrastructure was inserted into the mix to handle the post-processing of the GA data.

They also mention the concept of assigning value to the assist visits, but this is more completely thought through in C3 Metrics' whitepaper.  

Overall, I think the guys at SEOMOZ have done a great job in starting to unravel the intricacies of crediting multiple sources and how that might work with Google Analytics.  And, honestly, I'm thinking about how to start working some of this into my own clients' analytics.  However, there are a lot of assumptions going on here, some untested theories, some early summer dreaming, and implimentation in a production environment seems pretty far away.  

But then, isn't that how everything great starts out?

No comments:

Post a Comment