I was recently running some reports for a client on the performance of their traffic sources using the new SiteCatalyst Marketing Channels report. I wanted to get the % of visits from each traffic source so ran the Last Touch Channel report using visits as the metric. This report was then exported to excel so that I could calculate the % visits by channel and create a chart. However I was a little confused when the sum of the percentages added up to more than 100%.
I checked the reported total for visits against website visits for that time period and the numbers matched up. However the sum of visits by channel was greater than this total. Using Discover, I tried running the report with two metrics, visits and entries. My expectation was that these two numbers would be the same as visits = entries (at website level) and each visit was attributed to a single traffic source. However this was not the case, the number of entries per traffic source was less than the number of visits (with no consistent pattern across the traffic sources). The reported total was basically the same for each metric but while entries summed up to their total, visits did not.
Interestingly, running the same report for First Touch Channel had entries and visits as the same by channel. The totals match those for Last Touch Channel reports but in this case, they were all correct.
After some thinking and feedback from Adobe, this behaviour can be explained and it actually is correct. Well all of it except for the total for visits on Last Touch Channel reports which I believe to just be plain wrong. Let’s use a very simple scenario to demonstrate:
- new visitor
- access site via an affiliate
- enter on homepage
- view 3 pages
- no purchase
- access site via organic search
- enter on category page
- view 4 pages
- no purchase
- Still Thurs, previous visit still going
- access site via paid search
- enter on product page
- view 8 pages
- place order
For the website as a whole, there will be reported two visits with 15 pageviews (3 during the first visit and 12 during the second) with one order having been placed. The homepage and the category page will each be reported as an entry page but not the product page as it was not the first page viewed during a visit.
When using the First Touch Channel attribution method, all of this visitor behaviour will be attributed to the first traffic source – Affiliates.
It is a different story when using the Last Touch Channel attribution method. There will still be two website visits but one visit for each of Affiliates, Organic Search and Paid Search. Each channel receives one visit as they were all the last touch point before at least one page was viewed. However there are still only two Entries, one for Affiliates and one for Organic Search. Paid Search does not receive an Entry as it was not the first page in a visit. The order is attributed to Paid Search though as the last touch point prior to the order being placed.
So the Last Touch Marketing Channel report doesn’t work quite as I thought it would, I had thought that each visit would be attributed to a single traffic source, the one that initiated the visit. Instead, because Marketing Channel rules are evaluated on EVERY image request, every new click through into the website is counted as a visit for that traffic source (sentence updated thanks to clarification from @OmnitureCare). The logic of how SiteCatalyst works with regards to visits, entries and the Marketing Channels report all makes sense but it does require you to think through what the numbers mean.
Further, this also means that any landing page analysis performed is not quite as complete a story as anticipated. Entry Page reports will only include landing pages that are the first page viewed during a visit, not necessarily every landing page occurance. This shouldn’t be a problem for understanding landing page performance as the sample size would be very large but it should be taken into account. Remember always, the purpose of web analytics is to use the data for decision making, not for reporting – numbers don’t have to be 100% accurate to be useful.