1.7K members online now
Understand information in your reports and troubleshoot reporting issues such as self-referrals, (not set) data, and inaccurate information
Guide Me

Content Grouping Reports Broken After Site Redesign

[ Edited ]
Visitor ✭ ✭ ✭
# 1
Visitor ✭ ✭ ✭

I launched a redesign of my website on June 6. My Google Analytics tags were not modified or changed, and most of my traffic reporting appears to be unaffected.


However...  Custom Reports which filter traffic based on Content Grouping appear to be broken, showing a sharp decline in Users after the site redesign was launched:


Screen Shot 2017-07-03 at 5.11.58 PM.png


This sharp drop in traffic was not reflected in any other reports. Indeed, looking at the other metrics in this same report shows no significant change. Here are the Sessions and Pageviews metrics from the same report:


Screen Shot 2017-07-03 at 5.11.23 PM.pngSessionsScreen Shot 2017-07-03 at 5.12.40 PM.pngPageviews

As you can see, these metrics show no major change, and both reflect the sharp traffic increase that we received starting on June 11.


Combining "Users" with other metrics using the "vs." feature also creates strange effects. If I compare "Users" with "Pageviews", for example, you will see the line graph for Pageviews shows a similar drop. However, this graph is simply wrong, as you can see from the dedicated Pageviews report above.


Screen Shot 2017-07-03 at 5.21.07 PM.png


Can anyone explain why this happened? Once again, my Google Analytics code has not changed — only the layout of my website. I have also tested my implementation using the Google Analytics debug features and confirmed that my Content Groups are being recorded as expected.


EDIT -- More Information & Clues


I have performed more troubleshooting on this issue, and I remembered that I did make one change to my Analytics settings on the same day as the site redesign was launched: I edited the "Default Page" field in the View Settings page for my site. It had (incorrectly) been set to "index.php" for years and I was tired of seeing that appended to all of the URLs for my site, so I just removed it. I do not understand how it could have such a severe impact on the "Users" metric in a Custom Report...


I have also discovered something very interesting: the number of Users reported for each day changes SIGNIFICANTLY based on the selected date range. Allow me to demonstrate...


If the selected date range includes the date I updated my site, June 6, a sharp decline in Users is reported. Note the highlighted number of visitors on June 8 in both of these graphs:


Screen Shot 2017-07-03 at 9.27.27 PM.png


The above date range of 10 days (Jun 1-10) with a sampling of 68.77% of sessions indicates that 5,604 Users visited on June 8. Now, if we change the date range to the ten days following June 6 (Jun 7-16), we have a remarkably different report:


Screen Shot 2017-07-03 at 9.28.12 PM.png


With this date range, it shows that 16,699 people visited the site on June 8 — that is a 198% increase achieved by simply adjusting the date range. (I know the second report is based only on 16.98% of sessions. To remove differences in sampling from the equation, I can restrict the date range June 8 itself to receive 100% sampling for the date; the true number for the day was 16,752.)


Something isn't right here...

Content Grouping Reports Broken After Site Redesign

Explorer ✭ ✭ ☆
# 2
Explorer ✭ ✭ ☆

The way you define your content grouping is up to you - I strongly suspect you are using regular expressions to categorize your content and some of those aren't matching anymore.

Content Grouping Reports Broken After Site Redesign

Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭

Hi Stéphane,


I am using regular expressions to filter my Content Groups in the Custom Report (specifically, I am using ^News to show the content groups that begin with "News:", as you can see in the reports above). However, none of the Content Group names were changed, nor were there any changes to the site structure — the redesign was purely cosmetic. The other Content Groups that I track are also experiencing the same mysterious drop in users.


Please see the new information I have added to my original post above. It shows that the data itself is still being recorded as expected, but it is not being correctly processed in the reports.

Content Grouping Reports Broken After Site Redesign

Explorer ✭ ✭ ☆
# 4
Explorer ✭ ✭ ☆

This is definitely a tracking or configuration issue - certainly not on GA's side. There's certainly something - is the number of pageviews consistent? Have you enabled the new user calculation option? Without access to your site or config, it's hard to pinpoint what might be broken.

Re: Content Grouping Reports Broken After Site Redesign

Visitor ✭ ✭ ✭
# 5
Visitor ✭ ✭ ✭

How can you claim an issue is "certainly not on GA's side" when we can introduce User reporting variations of more than 202% by adjusting the date range?




In addition to the discrepancies I documented in my original post, it's worth noting the above graphic indicates this bug is affecting traffic that was recorded before I made any changes to my site or GA account settings.


Also, yes, as reported in my post above (with screenshots), no other Metrics appear to be affected — the number of Sessions and Pageviews both appear normal and consistent. Likewise, no other reports show any drop in Users across the site which would be expected if this were an implementation issue. Only Custom Reports with Filters show the sharp decline in traffic. I have verified this by filtering other Dimensions (not involved with Content Grouping) and the same mysterious User reporting traffic drop appears.


Content Grouping Reports Broken After Site Redesign

Explorer ✭ ✭ ☆
# 6
Explorer ✭ ✭ ☆

Sorry, I didn't meant to be rude when I said that. In my experience, virtually every single time someone claimed a "GA bug" it turned out there was a very good explanation. One thing I notice in the 1st screenshot is "this report is based on 13.84% of sessions" - this is pretty high sampling and likely makes those results totally unreliable. Have you tried showing New Users or Returning Users as a secondary dimension? If anything, I would trust the results without any sampling...