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.


Friday, July 23, 2010

Test Results: First vs. Last Touch Keywords with Google Analytics

The idea of what the value of topical search is in a situation where branded phrases clearly dominate the conversion metrics is not a new one.  This is a big motivator for multi-touch tracking.  I wrote about SEOMOZ's article on multi-touch keyword tracking previously, and this got me thinking through how you would actually implement first and last touch tracking in the current Google Analytics platform.  So, I made some changes to the tracking script and pushed it live. 

The results of my first and last touch keyword attribution test were not as robust as I'd hoped.  The usable data set was much smaller in the end than I was looking for. After filtering out all the things you can't use, you find this type of analysis really applies only to a small group.  Mostly this is becuase I was primarily interested in the cases where the first touch keywords as different than the last touch keyword.  The ones that were the same were not as interesting to me.  that being said, seeing the difference in volume was eye-opening.

A custom Google Analytics tracker was created and placed on a website for 30 days.

The test was to see:

1. How often visitors returned to the website on a different keyword.
2. How often topic initial keywords led to brand keywords on repeat visits
3. How first vs. last keywords differed in general.

During that time, 860 visits were tagged from google search with a custom variable that captured the first visit's keyword. 57 visits were scrubbed for weirdness.

459 visits were recorded as 100% new and were filtered out.

After filtering 100% new visits and funky data, the working data set was 344 visits.

264 visits were tagged as having identical first and last touch values.

80 visits showed different first and last touch keywords. (23%)

Of these, 61 showed a branded first touch keyword and 19 showed a non-branded keyword.

Branded First Touch Keywords
19 out of 26 distinct first touch keywords were some kind of branded keyword with many mispellings. Of these, the second touch keyword was set by google analytics as "not set" in 16 cases. In 100% of the remaining cases, the second touch keywords was a branded phrase of some sort, usually very related to the first touch phrase.

Stacked bar showing the volume of visits relative to each other.
The visits that were most relevant to this test turned out to be a very small percentage of the total.

Non-branded First Touch Keywords
In the remaining cases, "not set" as the last touch keyword became more common (12 out of 19). In the remaining 7 cases, a little more than half (4) had a branded last touch keyword.

Summary and Next Steps
Overall, when thinking about how topical phrases and brand phrases interact with each other, you're really talking about a strick minority of the overall visits to a website. This perspective raises the question of how much this type of analysis really worth at the end of the day. But if you are really interested in web traffic behaviour like I am, you do it anyway because:

1. Finding out that people are most likely to keep finding you in the ways they have already found you is interesting and helpful. They are unlikely to change the way they think and find you.

2. Understanding that if you do want to change the way they think and find you, the path of least resistance is moving from topic to brand, not from brand to topic. No one who used a branded phrase returned on a topical phrase, but the other way did show some movement. This can be important for how you position a product in marketing because, in my mind, this represents more of a bottom up approach, rather than a top down approach.

Next steps are really to get a bigger data set. Reliable conclusions are hard to claim when looking at such a small group. Also, I want to factoring in conversion rates to determine traffic value would add more context. So, more testing is on the way!

No comments:

Post a Comment