Analytics
7.1K members online now
Understand information in your reports and troubleshoot reporting issues such as self-referrals, (not set) data, and inaccurate information
 
Guide Me
star_border
Reply

Individual Content Group Usage Sums to Greater than Total Pageviews

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

Alright... this is confusing. I've searched for answers, but can't seem to figure this one out. I've got a Custom Report that looks at two Custom Content Groups, but the overall pageviews are in some places greater than the individual groups put together!

 

Any thoughts?

ga-sumthingwrong.png

I do have a filter on the report excluding (not set) pages, but that should exclude from all counts, right?

 

Thanks for any suggestions.

1 Expert replyverified_user

Re: Individual Content Group Usage Sums to Greater than Total Pageviews

Top Contributor
# 2
Top Contributor
Hi Scott,

Double-check your content-group definitions by adding 'Page' as secondary dimension in this report. Be sure to have included the (not set) pages for testing because they may actually contain missing definitions. As websites get updated and modified you should review your content-group definitions on a regular basis.

Christoph

Re: Individual Content Group Usage Sums to Greater than Total Pageviews

Visitor ✭ ✭ ✭
# 3
Visitor ✭ ✭ ✭
OK, I tried that, but didn't help. I was excluding the (not set) pages intentionally as I didn't want that cluttering up report. Those millions of pages wouldn't have this parameter set anyway.

The Content Group codes are injected based on the site's CMS use of a template And after checking with Tag Manager and Dev Tools Console, I can see that the correct codes are being sent. (Using "Page" as a secondary dimension didn't change anything.)

I also tried removing the higher level drilldown.

This is just exceedingly odd to me as I don't see how one component can possibly be greater than all pageviews of the constituent parts. It's just bizarre.

Re: Individual Content Group Usage Sums to Greater than Total Pageviews

Top Contributor
# 4
Top Contributor
Hi Scott,

okay, I see. I'm wondering if it might be a sampling issue - is there data sampling applied? Do you see the yellow bar in the upper right? Does the data change if you shorten the time range, e.g. to May, 10 only?

Christoph

Re: Individual Content Group Usage Sums to Greater than Total Pageviews

Visitor ✭ ✭ ✭
# 5
Visitor ✭ ✭ ✭

 

 

 Yup. There's absolutely sampling going on. And ye... data certainly changes when I shorten time range. So definitely good thoughts.

 

Still, even scoped down to one day, the two sub-parts, the numbers are off. For example, I can scope to one day, and the two sub-parts don't quite add up to total pageviews. And yet, when I expand the range to several months, I can see some days that have the issue. So maybe I'm just stuck with this as a sampling issue and I can only use this data directionally. Watch this example of a particular type of content using Content Groups breakdowns...

 

Here's just one day when the dates are scoped to several weeks...

 justoneday.png

Now, here's that same day, scoped to just that one day, June 1,

sameday-onedayscope.png

Same total, different sub-part numbers. The sub-parts should never really exceed total pageviews though. So it's all very strange. I'm not sure what to trust. I can always go back to individual page views for the pages to see what's going on in standard reports. (They kind of don't match either.) But in this case, I'm trying to judge based on these content types. I had to use Content Groupings because the URL paths don't break out the types.

This is actually for a consulting client. I'm not sure what to tell them. (Other than buy the Enterprise version or use another product.) If I can fairly tell them, "look, it's just sampling and it happens a bit inconsistently," and that's True, then ok. And I think it still tells them generally what's going on; it's just not something they can use with precision.

 

It just does seem like rather inconsistent behavior.

 

Thanks for your insights,

Scott

 

 

Re: Individual Content Group Usage Sums to Greater than Total Pageviews

Top Contributor
# 6
Top Contributor
Scott,

yes, that's definitely confusing ... Is your data sampled even for a single day (yellow bar)? Remember that standard reports (i.e. everything without any filters, segments, secondary dimensions or custom reports) always show completely unsampled data. Where did you actually filter out the (not set) data?

In a chunk of my own data, I see unsampled data for both 'Page' and various Content Groups when applied as primary dimension without any further filters or segments etc. even for more 1 million pageviews. Only if I apply a secondary dimension for example, sampling enters.

Christoph

Re: Individual Content Group Usage Sums to Greater than Total Pageviews

Visitor ✭ ✭ ✭
# 7
Visitor ✭ ✭ ✭
Yes. Even one day has sampled data. I'm applying the (not set) filter directly in the custom report. In this case, there's a Content Group called Quiz Types. And when there's an exact match for "(not set)" I exclude them.

When I remove the filter, it sort of fixes the problem in that the sub-parts no longer exceed the whole, but the issue here is that even for a day I've got several MILLION not set entries. (it's a high traffic site.)

If I check an individual page, (that is just the standard Page, which is the URI), there's still a discrepancy between the Custom Report results and when I go into the standard "All Pages" report. The Standard Reports consistently shows higher numbers. Not only that, but if I do as you did and apply the Content Grouping on a Standard Report, the numbers go down a bit, but not quite to the Custom Report Level!!! Madness!!!

So it seems like the answer is threefold:

1) It's a sampled data issue.

2) The Custom Reports are not precise. They may be fairly used as directionally accurate guides for which content types are doing best as the sampled data should be at the same sampling rate across types.

3) The standard reports will show page views differently, (generally higher), than the Custom Reports because they're not sampled. And even if you apply the single primary dimension to a Standard Report, the numbers go down as some form of sampling starts occurring, though perhaps not to the degree a full Custom Report will experience.

This is all kind of lousy of course, but at least we can interpret what we want to know if this is all true. To get real numbers, we'll have to pay up big for a larger solution or use another tool to figure out these pages alone.

I appreciate all your insights here and hope that I've described things properly. Does what I've said seem true? It's one thing to make a mistake. It's quite another to be false and I don't want to tell anyone else the wrong things.

Thanks,
Scott