Many people might assume that some very basic data that gets reported through web analytics tools are measured the same way. For example, visits, unique visitors, page views, conversions (however you define yours), revenue, etc. However you might be surprised to find out two data points that are reported very differently in Adobe Analytics and Google Analytics.
Time Per Visit
At first glance, this seems like a pretty straight-forward metric, right? Nope. What happens on the last page you visit on a site? How do the systems know how long you stayed on that page? The answer is that they don’t. And therein lays the problem. Both systems will report your visit time as the time you enter the site until the last click made on the site. It does not count any time spent on that last page, since it is unknown. But what happens if you only view one page? There is no second click to measure your end time. Here’s where the difference is between these reporting systems. Adobe does not count those visits, since the time spent is undetermined. Google reports 0 time on site. This can lead to some pretty different averages if you are comparing reporting tools.
Let’s look at this example:
Visitor 1 spends 3 minutes on your site
Visitor 2 spends 1 minute on your site
Visitor 3 bounces
Adobe reports an average of 2 minutes (4 minutes total divided by 2 visits).
Google reports an average of 1.3 minutes (4 minutes total divided by 3 visits).
The key takeaway here is that neither are absolutely correct. We just need to understand the differences between the tools, and be looking at trends over time. The number, in and of itself, can’t be used without some kind of context in which to examine its meaning.
How would you define direct traffic? My definition is any traffic coming to your site where a referring domain cannot be detected. Again, this appears that it should be a no-room-for-discussion topic. However, we came across a large discrepancy between Adobe and Google when it came to natural search referrals for a client. Google showed a much larger percentage of traffic from natural search and smaller percentage from direct traffic. Overall, the totals were comparable.
We tried to narrow down the discrepancy. It simply seemed that one system was either overstating or understating the attribution to natural search. Why would this be? I contacted support at Adobe to see if they could help us. The only thing they could do was tell me how SiteCatalyst natural search attribution works, which I already knew. I really wanted someone to look at our code to see if something wasn’t implemented correctly. They offered to send us raw data feeds instead. We used the raw data to look at 50 different test visits we had our intern make to the site, using different keywords, browsers, and operating systems to get there. We were able to verify that all 50 came through. Oddly though, they also came through Google Analytics correctly. What were we missing? How did all of these visits come through correctly in both places, and yet Google shows double the natural search visits than SiteCatalyst? We tested a different site using the same shopping cart system with a SiteCatalyst implementation to see if it was something in the shopping cart causing the issue. Nope – the discrepancy didn’t happen on this other site.
I searched the help files in SiteCatalyst and Google Analytics, I tweeted, I posted in my LinkedIn forums…no help. Finally, I found a tiny blurb in Google’s support site on campaign tracking that says, “If it is the user’s first visit to your site, the tracking code adds the campaign tracking information to the cookie. If the user previously visited your site, the tracking code increments the session counter in the cookie. Regardless of how many sessions or how much time has passed, Google Analytics remembers the original referral.” OK, this sets me on a wild goose chase to find further information about the various cookies used to track campaigns and sources. On developers.google.com, I finally found what I was looking for: The _utmz cookie stores the referral data, whether it be a referring domain, campaign, or search engine. The cookie expiration is 6 months. 6 months?!?! I continued my search for more info. I found a Yahoo user forum that posed the same exact question I had. The kind person who answered specifically said that Google will cookie your machine upon your first visit to the site. That source will remain until another source/campaign is available to overwrite the original source. Direct traffic is NOT considered a source. Holy cow!
What was happening with our client is that because they have such a high number of repeat visitors, most within that 6 month window, that their return traffic was showing up in Adobe as direct, and in Google as search, since that was their last “known” source.
You can now change the default setting of that cookie to something other than 6 months, but why would it default to that length? That is a VERY long time. Let’s not forget that if a visitor returns to the site within that 6 month time, the cookie expiration renews. Why not set it at the same length as AdWords? Secondly, why does Google not see direct traffic as another source?
Let’s take a look at an example from a site like amazon.com to see the implications. A visitor searches Google for the term “baby cribs.” The visitor clicks on a search result that goes to one of Amazon’s baby crib pages. The visitor buys or doesn’t buy something from Amazon at this time; it’s irrelevant for this example. Tomorrow, the same visitor decides that she wants to purchase a book she’s wanted to read. She types in amazon.com in her browser, looks for her book, and purchases it. That book purchase revenue and visit is going to be attributed to the “baby crib” keyword. This is obviously an extreme situation; however, it does make my point. The attribution of that second visit truly was a different source than the first. It certainly would have had different traffic patterns on the site. Yet, all of that was attributed to the initial keyword the visitor came through because there was nothing to override that source. It’s just ludicrous to me.