Earlier this year, Digital Initiatives started implementing Google Analytics (on January 6, 2014). I thought at the time that switching to this program would make a great blog post: I’d have plenty of time to install the code, work through the CONTENTdm tutorial, and then have a good month to tweak any settings and report back to the group. Since then, I’ve realized it’s not going to be that easy.
First, let me explain how we harvest statistics. Bill Yungclas is responsible for the CONTENTdm part of our stats, and I am responsible for the offsite web pages. CONTENTdm has an in-app stat program (including item views per collection, items views by country, items in collection) and I use StatCounter (including page views, unique visits, first time visits, returning visits). StatCounter is easy to set up, and works like a charm. CONTENTdm calculates compound objects using a different formula than Google Analytics: “…total views per object are displayed, as well as the number of views per page.” OK, so it sounds like they track compound objects differently, and that those “pages” might skew upwards. If accurate, this is probably a good reason for CONTENTdm to move away from in-app stat tracking and recommend Google Analytics; the latter tracks each page clicked on. So, there are issues between how CONTENTdm and Google Analytics track: primarily, the stats are inconsistent.
Bill and I did an experiment, and it appears Google Analytics is correctly tracking each page. However, in the short overlap between the two tracking apps, the counts aren’t anywhere near being equal.
Herein lies the issue: is CONTENTdm bloating the stats, or, is Google Analytics not counting correctly? Since our experimentation with page clicks appears to identify that the latter is working correctly, then, we have to imagine that CONTENTdm is indeed skewing the stats. An example: CONTENTdm tells us that total items viewed for January 2014 was 62,453. Google Analytics says “nuh-uh;” it’s really 896 visits to our site with 3,984 individual page views. Yikes! 62,353 vs. 3,984 works out to +/- 58,369 difference in views. So what exactly is CONTENTdm counting and what does an “item view” constitute for them? I’d like to say the 3,984 page views from Google Analytics is off, but considering that StatCounter lists 536 hits for our offsite home page, and Google Analytics is saying it tracked 475 during the similar period for the onsite home page, I have to think that Google Analytics is more accurate. Suddenly, my nice little blog post has expanded to one that needs further investigation, insight, and time.
Beginning in March, I’m going to begin tracking our offsite web pages with Google Analytics to see how the stats compare to StatCounter. We’ll have to wait to see what a full month using Google Analytics does to our numbers, and I have a feeling that the count isn’t going to be anywhere near the 60,000 range. We are hoping to stop using CONTENTdm for tracking (Bill was told it’s being discontinued, and it could be any day) but we will have to do some soul-searching in the meantime.