I've been using Confluence (Data Center) to host a company-internal Blog. It works great in many ways. But frankly, the analytics data looks just plain wrong.
I tracked analytics by page (topic) over time, adding it to an Excel chart to visualize:
This is obviously super useful to track engagement, relative interest in topics, etc.
Some of the data was a bit manual. For example, to get the "Unique Reader View" values above, I'd go to the Analytics page for that topic and simply count the number of viewers listed (excluding myself) - so 3 show below:
However, observe how the bar chart data above drops off. The right-most bar for the first three topics significantly decrease. What's going on?
Analytics data is only saved for a year, which I didn't realize. As a result, this way to track engagement fails over time. Sadness.
So I explored other options in the Confluence analytics data, including some line charts.
The following monthly roll-up data looked valuable in an on-going way. This is allegedly showing unique page views by month:
Excel can easily aggregate these "unique reader page view" counts, giving me an ongoing total. But this data doesn't match up:
On the left are the manual counts of all the listed "page readers" over time for one of my pages. (I tracked data more frequently early-on for a topic, and less frequently as topics aged, to reduce the effort)
The "Monthly Delta" numbers come from the "Unique" data shown in the "Content Analytics" screenshot above. All this data is for the same page over time.
The "Accum" column simply aggregates the Deltas. As you can see, there is NO WAY that the "delta" data aligns. From Mar-11 to Aug-11 there were 10 new names listed on the page. But the Deltas over that time period total between 15 and 20!
From Dec'22 to Aug'23 the two sets of data are off by over 100% (ten more vs 23 more).
This can likely be reproduced in your instance quite easily. Find a well-read page that's maybe 6-12 months old. Count up the number of reader names. Then show the "Monthly/Unique" chart and total up the deltas. Do they match? I expect not.
The documentation doesn't help explain any of this.
I'd like to be able to trust the page analytics in Confluence. But given the above analysis, it seems I can't.
Can anyone clarify what's going on here? Someone at Atlassian might want to consider reaching out to me to reproduce these results, and perhaps investigate the inaccuracies in the data I'm finding in Jira and Confluence. Verification by others also would be appreciated - do you see this in your Jira and Confluence?
P.S. Please note that I am not interested in Marketplace apps to help here.
P.P.S. If you total up the monthly values when Display="Total" (rather than "Unique") you'll see that it exactly matches the sum of the "Views" in the detail list below. So some of these analytics correlate -- but the Unique counts are not even close. Why?
P.P.P.S. I had reason to investigate the data in the Jira "Control Chart" report. Some very simple testing shows errors of 7-8% in the Cycle Time data.
I'm reviewing some document analytics today and also finding that things don't match up. Did you ever find any answers or solutions?
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.