Analytics
4.7K members online now
4.7K members online now
Dive into multiple domain (Cross/Sub) tracking, implementing Ecommerce and Enhanced Ecommerce, setting up Event tracking, and Universal Analytics code.
 
Guide Me
star_border
Reply

Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

[ Edited ]
Follower ✭ ✭ ☆
# 1
Follower ✭ ✭ ☆

I'm migrating a high-traffic site from Classic Analytics (ga.js) to Universal Analytics (analytics.js). Both setups have been running in parallel for the past few days: ga.js is hard-coded on the pages (i.e. "_gaq" calls) and analytics.js is running through Google Tag Manager (GTM) with the standard integration for pageviews. Furthermore, both GA properties and views are configured exactly the same, and there are no interactive events (or similar) that could be causing big discrepancies.

 

There are some really strange discrepancies in the data, which I can't fully explain. For example, looking at yesterday's numbers, analytics.js reports:

 

  • 4% more sessions (~5,000)
  • 1.5% more users (~1,500)
  • 2% less pageviews (~7,500), and mostly flat unique pageviews
  • 6% less pages/session (2.69 vs 2.85)
  • 10% lower avg. session duration (00:02:31 vs 00:02:48)
  • 2% higher bounce rate (50.4% vs 49.4%)

 

I started drilling down into the reports and, so far, the biggest differences I can find in the analytics.js property are:

 

  • 15% more Direct sessions (~3,200)
  • 115% more sessions from "UC Browser" browser (~3,500)
  • 130% more sessions from "level 3 communications inc" ISP (~2,100)
  • 320% more sessions from the city of Putney (~1,700), with the majority of the increase attributed to "UC Browser"
  • 200% more sessions from Dubai and Beijing.

 

While I know analytics.js is supposed to be better, I'm having a very hard time explaining such large discrepancies, especially when it comes to the higher sessions, lower pageviews and time on site: these are moving in contradictory directions.

 

I would really appreciate any ideas on the following:

 

  1. How is it possible that I'm seeing so many more sessions, but less pageviews?
  2. Why is avg. session duration drastically different?
  3. Are there any other factors that could be causing such discrepancies? Especially the inflated session number.
  4. Should I move analytics.js entirely out of GTM?

 

Any other ideas are greatly appreciated.

1 Expert replyverified_user

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Top Contributor
# 2
Top Contributor
I assume these are in separate properties with different tracking codes?

Based on what you have stated, I would say the old property is being spammed in a way that the new property/GTM is oblivious to. Continue digging.

A word on your metrics: ignore ALL the rate and percentage metrics for this investigation. They don't actually tell you anything. If your sessions are up and your pageviews down, the pages/session will be way down -- it's just a calculation. Same with avg session duration -- add a bunch of 0 sec sessions and the average drops. Your real questions need to be what pageviews dropped, and where are the increase in 0 second sessions coming from. Homepage and UC Browser? Then it's bot traffic.
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

[ Edited ]
Follower ✭ ✭ ☆
# 3
Follower ✭ ✭ ☆

Thanks for your reply. Yes, the properties are tracking under different tracking codes.

 

I looked at sessions with 0 second duration and, while the discrepancy holds between the two trackers, it doesn't account for the entire session difference. The same applies when I look at sessions with a single pageview. I also looked at sessions with 0 pageviews, and analytics.js reports 1 and ga.js reports 11. So neither of these cases seems to explain the problem.

 

Now, if I create a segment to exclude any sessions that are Direct source or city Putney, Dubai, Beijing or "UC Browser" or "level 3" ISP, the analytics.js property shows 2% less users and 7% less pageviews than ga.js.

 

So, at the top-level, I'm seeing a lot more sessions and less pageviews, but when I exclude the "questionable traffic", the pattern reverts and I end up with less sessions and less pageviews.

 

One theory is that analytics.js is malfunctioning and is (a) logging less pageviews, and (b) splitting sessions for certain types of traffic, causing this "session inflation".

 

Do you have any other ideas that might help me find the root cause of the issue?

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

[ Edited ]
Follower ✭ ✭ ☆
# 4
Follower ✭ ✭ ☆

Another thing I just noticed: Looking at the User Explorer report in the analytics.js property, I see that users have a lot more sessions/user than what ga.js is reporting. Here's a screenshot of one such user from yesterday: notice how new sessions get created only a few minutes apart, significantly under the default 30-minute cut off.

 

Is it possible that something is wrong with how analytics.js is handling sessions? I would expect the User Explorer reports in analytics.js and ga.js to line up, but this is not happening...

 

 Screenshot: http://imgur.com/a/C8KmC (image upload isn't working in the forum for me -- I get an invalid HTML error when trying to upload the image).

 

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Follower ✭ ✭ ☆
# 5
Follower ✭ ✭ ☆

@Mike S: I did some more digging and found something interesting...

 

Using a custom segment, I was able to isolate a single user in a single day and the results are striking: the analytics.js property shows 50 pageviews and 20 sessions, while the ga.js property shows 50 pageviews and 8 sessions. When I look at the Channel report, analytics.js is showing 13 Organic Search and 7 Referral sessions, while ga.js shows only 8 Organic Search sessions. So it looks like analytics.js is heavily splitting sessions.

 

Do you know what could be causing this?

 

Thanks again for your help.

 

 

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Top Contributor
# 6
Top Contributor
Look at the landing pages and source of the differing sessions for further clues.
Check the Referral Exclusion List on both properties. You may find that one of them is splitting transfers to a payment processor from the returns and the other isn't.
Check the actual referral sources - are they from your own site? If so, the problem might be the session duration specified in one property is shorter than the other.
Finally, check your GTM configuration - are you doing anything that could force a new sessions?
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Follower ✭ ✭ ☆
# 7
Follower ✭ ✭ ☆

@Mike S: Thanks again for the reply. I did some more digging, and I can confidently say that there are no differences in configuration (referral exclusions, session duration, etc): the settings are exactly the same, with the only difference being that the legacy property runs with hard-coded ga.js, while the new property runs analytics.js through GTM.

 

In my latest research and tests, I was able to isolate two things happening:

 

1. analytics.js is creating new sessions for Referral sources that ga.js is not

 

Specifically, when a user goes from Direct to Referral or Organic to Referral, the analytics.js property creates a new session, but the ga.js property does not.

 

In the earlier case of the user that recorded 20 sessions in analytics.js but only 8 sessions in ga.js, I can see that these extra sessions are the result of the user (a) first arriving from Google Organic, (b) leaving the site, and (c) returning (within 30 minutes) through a partner site that links to my site. This immediately causes a new Referral session to get started, and end the previous session.

 

2. But analytics.js is also creating new sessions for single-source users

 

For example, I isolated a single "bing /organic" user: analytics.js recorded 15 sessions and ga.js recorded only 8 sessions. New sessions are getting created even if the user never changed traffic source and has hits within 30 minutes of the last interaction. Here are two screenshots that show exactly the issue:

 

 

 

All of this leads me to ask...

 

  1. Is there a known difference in how ga.js and analytics.js handle Referral sessions that explains what I'm observing?
  2. How is it possible for analytics.js to create new sessions when the user is browsing from the same source, is within 30 minutes of the last hit and is not browsing past midnight? I can't think of anything else that would cause these additional sessions.

 

Again, thanks for your help and I would appreciate any other thoughts.

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Top Contributor
# 8
Top Contributor
Interesting..... your #1 scenario, analytics.js is working the way it is supposed to, as I understand it, splitting the session to the new source. It looks like ga.js didn't work the way it was supposed to.

Your #2 scenario, I don't think that should be happening, unless the user returned to search and came back from another search result -- same source, different referral instance?
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Follower ✭ ✭ ☆
# 9
Follower ✭ ✭ ☆

Thanks @Mike S.

 

For the #2 scenario, I think you're right! I just looked at this user who arrived from non-SSL "bing / organic" source and, since the incoming keywords are exposed, analytics.js picked up 15 unique keywords, while ga.js picked up only 8. This would explain the difference in sessions.

 

For scenario #1, I just tried reproducing the case of going Organic to Referral and, when I observe the "utmcsr", "utmccn" and "utmcmd" values being passed in the ga.js and analytics.js calls, I no longer see the Referral in there: it continues to pass the "organic" source. I will check in GA tomorrow once the activity has been processed. Because I would expect at least analytics.js to be receiving the referral information to generate a new session, as it did yesterday.

 

Is it possible that the session splitting based on source is not always functioning properly in analytics.js and ga.js? Since it looks like it doesn't always send the correct values, even under exactly the same circumstances.

Re: Why does Universal Analytics show more sessions, but less pageviews than Classic Analytics?

Top Contributor
# 10
Top Contributor
I am not the guy to ask, but I do know that there are very big differences in the old vs the new, so I guess I am not surprised that there are some subtle differences in how a few scenarios are handled. I'm also not sure if this should cap your investigation. I guess a bit more validation before you make your conclusion is in order. Good luck...
Mike Sullivan, Google Analytics Top Contributor
Was my response helpful? If yes, please mark it as the ‘Best Answer.’ Learn how here.