Why there is huge difference in Pageviews tracking number on discourse admin dahboard and on google analytics

Anyone else seeing big discrepancies between Google Analytics “page views” and API Calls on the dashboard these days? Particularly on our hosting or self hosting?

1 Like

Digital Ocean hosting, tests-passed

9,789 Api Requests in last 7 days
(1,466 crawlers, 4,891 Logged in, 3,432 Anon)

6,532 Pageviews

Not orders of magnitude like the OP, but still a 50% increase from the google figure. The proportions hold up historically - google consistently reports ~1.5x as many page views as API calls.

4 Likes

We are seeing very big differences here:

The ‘pageviews’ according to Discourse are 65K, but according to GA they are only 13K.

So something is not right here.

I am working on reconciling this at the moment and am adding more internal diagnostics.

In particular I will ensure we set a header on every response that generates a page view.

Off the bat a big discrepancy is that analytics does not count web crawlers or anything that does not run JavaScript

3 Likes

Same here. For one day I had:

Google Analytics //

  • 163 Sessions
  • 61 Users (Logged and Anonymous)
  • 1 863 pages views

Meanwhile the dashboard report:

  • 4 800 Pages Views

16 from webcrawler an bots
2 500 from logged users
2 300 from anonymous

2.5x times more. That will put me in the 140k PV per month and the site is not launched yet. I am quite worry I will bust the 650k by far.

Since my original message I have reconciled multiple sites multiple times.

We simply count differently to what google does.

Google uses JavaScript ajax calls to count page views. We work hard only to issue them on first load or when you transition routes (like move from topics to latest list)

Our Page View counting is performed on the “raw requests” made (accounting for a header that informs us when to count)

Bots, JavaScript disabled are counted by us but omitted from Google.

11 Likes

It is worth mentioning that we added better bot user agent reporting plus user agent blacklist and whitelist support in 2.0 to deal with badly behaved bots of which there are … many.

5 Likes

Has something changed recently? My Google analytics are showing a big drop in traffic but my Discourse stats do not reflect the same kind of drop?

We more aggressively throttle bots and crawlers in Discourse 2.0, particularly Bing which was pathologically bad. You can search if you are curious.

2 Likes

I am seeing the exact same thing…

May 2018

June 2018

July 2018

Hi @codinghorror, The thing is, my discourse stats have remained more or less consistent, it is my Google analytics which have dropped - at pretty much the same time as I added the advertising code. Wondering whether it is possible my ad code is interfering with the Google Analytics somehow.

@jerdog did you add any new plugins or code to the site? Seem to have recovered in your last screengrab?

2 Likes

The only plugin I added in that span of time is the GitHub - discourse/discourse-github-linkback: Links github content back to a Discourse forum

It has begun to normalise again starting a few days ago which makes me wonder what happened to the codebase that might have contributed between June and late Sept?

Does anyone else see the same thing?

To me it looks like you had some sort of bug going where zero traffic was reported to analytics. I know we recently reorganised how we do analytics so it is very possible you still have the bug going but our new design works around your bug.

3 Likes

Hmm - ok I’m not seeing that across the rest of our properties so it’s a bit strange.

Not having checked my analytics for a while, it seems the issue is fixed since early August, so possibly fixed by the change you suggested above! nice one! :grinning:

3 Likes

It looks like Discourse is reporting a pageview to GA on only the first load of the page, but is internally tracking a pageview every time a new page of content is loaded inline.

It seems like it’d be more appropriate for Discourse to report a pageview to GA every time the URL updates.

For example: load a discussion, URL is /t/topic/1000, scroll to the bottom and the next page of posts comes in, URL updates to /t/topic/1000/20, and a new GA pageview should be logged with that URL.

2 Likes

You mean the first page of a topic (the first 20 posts) ? Yes, that’s how it works.

2 Likes

Yes, it looks like Discourse is tracking pageviews internally differently from what it’s reporting to GA, which is why the GA numbers look consistently lower.

It would be great to fix it to send a new pageview to GA every time the URL changed.

No, that isn’t why the numbers are different. This is why:

Half of the people browse with Adblock blocking Analytics nowadays, and Discourse internal numbers can count those.

That be weird. We update the URL for every post, do you think reading this topic warrants 40 pageviews?

2 Likes

You’re right, updating on every post would be too much. However, loading in the next page of posts should count as a new pageview. (and using the URL that is in the address bar is a natural fit for this.) Otherwise we end up with pageview numbers in GA that are much lower than actually representative of forum activity.