Google Analytics, Web Marketing Campaigns, and Conversion Attribution

In the words of web analytics commando Avinash Kaushik, “I humbly believe that the world of data perfection (“clean auditable data”) does not exist any more.” And he’s right! In the interests of moving away from data regurgitation and towards drawing actionable conclusions in the sea of web analytics and statistics, the first step should be the realization that even today, with all the cool gadgets at our disposal, web analytics data is imperfect and inherently flawed.

I mention this not to provide an excuse for incorrect conclusions and strategy (“If only that tracking code was installed correctly on my blog, I would’ve known to avoid using flashing strobe banner ads!”), but rather as a step towards analytics enlightenment. Yes, data may be flawed, but this should be your call to arms – find out how the data is flawed (or, stated another way, the limitations of the data set) and what you can do, as an analyst, to draw real-world conclusions from these numbers.

Let’s look at an example. To let you identify high/low performing marketing efforts, Google Analytics will automatically track the source of visitors to your site. These are segmented into a few default mediums, visible in the Traffic Sources reports. The “cpc” segment captures traffic from “cost per click” or “pay per click” campaigns, such as Google AdWords. From here, you can see how many visitors that clicked through from your pay-per-click ad actually bought something from you, filled out a lead-generation form, etc.

But! You now know that this data has issues. Remember, your newfound acceptance of this axiom shouldn’t be an excuse, a crutch to lean on; instead, this should inspire you to dig deeper and not take these numbers at face value.

And here’s why: there’s a specific way in which Google attributes things like conversions and e-commerce transactions to traffic sources. Let’s say a user visits your website several different times via several different mediums, and finally buys something or completes a conversion. How do you know which of these mediums drove that purchase? With one exception*, in Google Analytics, conversions are attributed to the most recent campaign or medium by which that user arrived. For example:

Due to a recent break-in at her home, Jane Customer is shopping for broadswords. To begin her buying research, she searches for “home defense broadswords” on Google. From the results page, Jane clicks on a PPC ad for ADT’s line of security swords. She browses the site for a while, checks out prices, but does not make a purchase – Jane wants to check out some other brands and options.
After sifting through the highly competitive landscape of medieval residential defense products, Jane settles on ADT and clicks on one of their banner advertisements on her Yahoo homepage. She arrives on ADT’s site and makes her purchase.
In this example, it’s fairly clear that Jane’s first visit, via ADT’s pay-per-click campaign, was probably most responsible for generating this sale! However, the purchase actually occurred on her second visit, when she arrived via a referral site – her Yahoo homepage. Google Analytics will attribute this purchase as being generated by a referral visit from Yahoo…and all the while your PPC conversion rate sadly continues to drop, bit by bit.

Uh-oh. So how flawed is your data? Can you make a serious business decision regarding the effectiveness of any given marketing campaign, based off incorrect or incomplete information? (Or as Mr. Kaushik would say, the “known unknowns”!) In terms of severity, this can be a serious issue if you’re selling a product that’s fairly expensive, fairly complex, or otherwise has a high-involvement buying process – every time a potential customer returns to your site, whether to continue researching or to purchase, their previous traffic source data is being overwritten!

(*The exception! For repeat visitors that return directly via a bookmark or typing the URL directly into their browser, Analytics will attribute any conversions to the immediate preceding traffic source. In the example above, if Jane’s 2nd visit to ADT was from a bookmark or simply recalling the URL from her first PPC-sourced visit, the conversion would still be attributed to the PPC campaign.)

So for e-commerce sites with these high-involvement/long buying cycle products, or sites with a very high Visits-To-Purchase ratio, analysts may want to direct Analytics to ignore these secondary campaigns and attribute any conversions to the first campaign that directed the visitor to your site. With links under your control (such as banner ads, external blogs, or links posted on your company Twitter account), simply append the variable “&utm_nooverride=1″ to the URL. This will prevent the original traffic source from being overwritten, thereby preserving the origin of each visitor.

Is this solution perfect? No way. But this knowledge of inherent error is what should keep you nimble as an analytics commando. The idea that you can always learn more and refine your conclusions…this is what keeps us on our toes.

Comments are closed.