Analytics-Adsense Click Discrepancy: How is this possible?

Confused

[kkinfy]

I have integrated Adsense with analytics for my website. While looking at 'Behaviour-->Publisher Pages' tab, I can see discrepancies (sometimes huge) between clicks recorded in Adsense and analytics. For example, there have been extreme cases where there would be just 1 impression but 3 to 5 clicks with 0 revenue in analytics (that too from US visits on a competitive niche). Is this a sign of double click penalty. Or is this a sign of invalid activity or even a bug in analytics reporting? Even so, on some days the discrepancies are so big that analytics tracks up to 2 times more clicks than Adsense. Kindly help.

Also, can webmasters share the discrepancy magnitude if any?

[NickMNS]

Adsense data reported in Google Analytics is unreliable so don't pay too much attention to it. There can be all kinds of reasons, exactly what the reason that applies and if any is unknowable. My experience is the opposite of yours. In my case GA under-reports, mostly due to amp-ads and the use of Ad-Manager, but there are certainly many other reasons.

I use GA as guide to help diagnose the source of problems. So, clicks with no impression is certainly a bad sign that warrants further investigation. My guess is that this is coming from user from social-media using mostly "Android-Webview" browser or "Safari-in-App" these are not real browsers, they are pseudo-browser that allow companies like Facebook to show your website content while keeping users in the Facebook app. As these are not full featured browsers they can cause all kinds of issues with JS and rendering that appears to be a source of the problem with AdSense. Now is this the cause of the double click penalty? It looks like it may be related but there has been no "smoking gun" proving it.

I suggest that you refer to this thread, where this is discussed at length:
[webmasterworld.com...]
@Janvitos offers a few screen caps showing the magnitude of some his most egregious click without impressions.

[kkinfy]

@NickMNS I am a little relieved to see I am not the only one who is facing the issue. I read Google documentation on data differences and I found these lines

"Separate JavaScript code: AdSense counts a page view when the AdSense ad code is executed by a user's browser. Similarly, Analytics counts a pageview only when the Google Analytics tracking code is executed by a user's browser. Because they are located in different parts of your page, it is possible that one of these JavaScript snippets will load and the other will not."

Can analytics somehow affect AdSense tracking? Especially if Adsense scripts execute after analytics script, can there be under-reporting in Adsense?

[NickMNS]

Can analytics somehow affect AdSense tracking?

No.

[ronron]

From my experience, seemingly clicks in Analytics are clicks before AdSense applies their filters. The revenue is on par in Analytics compared with AdSense, it's just the clicks that are magnitudes off.

[kkinfy]

@ronron I am also thinking the same that the clicks in analytics maybe before filters. This means the clicks recorded in analytics and not in Adsense should be invalid clicks (by bots/humans). In human cases, the site design and ad placements will need a review. But in the case of bots is there a way to prevent at least to some extent such clicks? Or is there some kind of script that will disable Adsense if user clicks multiple times?

[NickMNS]

I just went and check my stats again. I now see what you are saying now. It was strange that I never noticed it before. So I went back in time, and checked my stats. I only have publisher pages data starting April 2019. At that time the numbers were off but not by orders of magnitude. I checked every few month, and up until November 2020 everything was fine. But on November 2nd the numbers jumped, and from that point forward clicks in GA are 5x to 10x or more off of what is reported in AdSense. Now this is only my account, but this seems very odd, as it coincide to the start of the "First Party Cookies" (October 16 2020) and it coincides with the steady rise of people complaining here of invalid click penalties.

@Ronron, kkinfy and others
Can you check your GA accounts to see if the dates coincide, the start appear to be November 2nd 2020.

[kkinfy]

@NickMNS I can confirm that in my case the discrepancy began climbing from November middle.

[ronron]

@NickMNS

Exactly November 2nd 2020 CTR increased 3x and remained that way. Interesting that it's happening to the rest of us as well. At least I know it's not something I may have done on my site to cause a surge in accidental clicks.

[kkinfy]

I can just think of three reasons in the order of probability (just my opinion).

1) A bug in analytics reporting.
2) Adsense has started employing too many filters, click confirmation boxes, etc.
3) There is an ongoing large-scale attack on adsense websites. (least probable- 0.001%, as someone from Google would have already noticed if so.)

@NickMNS If some big/premium publisher like you could talk with Google reps on this, learn and share the cause, it would benefit a lot of small and growing publishers like me.

[NickMNS]

@kkinfy

1) A bug in analytics reporting.

Unlikely that the problem lies with GA - this is an Adsense issue

2) Adsense has started employing too many filters, click confirmation boxes, etc

I doubt this has anything to do with filters

3) There is an ongoing large-scale attack on adsense websites.

Unlikely, as the issue is affecting many people (currently we have a sample size of 3) but not everyone is impacted with penalties. I doubt that this is an external to AdSense.

Everything is pointing to some kind of a bug with regards to the roll-out of this "First Party Cookies" feature.

[kkinfy]

@NickMNS By bug, I meant a bug in the integration logic/code between Adsense and Analytics. Maybe analytics tracks some events as clicks that are not actual clicks. My real worry is should we consider clicks shown in analytics as real clicks? Because, if those are real clicks, at least we can try to identify the source country/device/browser, etc. (I know google doesn't share IP details due to privacy concerns).

[NickMNS]

...should we consider clicks shown in analytics as real clicks?

I don't know. It is impossible to know if these are actually clicks happening, or whether there is some event that is triggered multiple times. I may muster up the courage to take a look at my raw logs a little later on, if these are clicks then one should see the page request appear in the logs. I have my doubts.

I'm bashing my brains against the wall trying to find bugs in my current work, so a change of scenery, reviewing logs, might help clear my head. I'll let you know.

[ember]

My Analytics clicks are sometimes four or five times what they are in Adsense. It's been going on since September or October, 2019.

I tried finding the extra clicks to no avail. I took ads off and put them back in different spots. Clicks kept coming. Waited for a penalty. None came.

Revenue between Analytics and Adsense almost always matches within a few dollars. I rarely get end of month clawbacks larger than 1%, if that, so I don't know if Analytics clicks are real and filtered out in real time or if they are phantom clicks or what.

I finally just quit worrying about the Analytics clicks. Nice to see that I am not alone, though.

[ronron]

My difference is pretty crazy. Yesterday for example, Analytics shows 23K clicks while AdSense gave me credit for 600 something.

[NickMNS]

I looked into my GA stats some more, look at only today's numbers.

Before I get into that just let me explain that my website is special in that I have many pages (millions) and that I get very few visits to any one page but many pages get single visits. Thus it is very rare for the same page to get an ad-click. One needs to take the joint probability of same page being requested and ad being clicked. In other words, the same page being requested may not occur frequently but it does happen,(say 1 in 25), now pair that with the CTR of 1% of all pages. 0.04 x 0.01 == 0.004 so 1 in 4000. I can refine it further by, narrowing down the page view by time of day down to the minute. So probability of click on ad, on same page, in the same minute equals almost 0.

What did GA find, every single page view with one or more ad clicks all occur within the respective minute that the page view occured. What this shows (in my case) is for a page that had say 7 clicks, those 7 clicks all occurred at the same time, this was true for all the pages. Moreover, the unique page-views that received at least one click coincides more closely to the number of clicks reported by AdSense.

I'm now moving to my logs

[NickMNS]

Some further digging. The affected pages are almost all AMP pages. So nothing on my server as the page view came from the amp-cache. I did find one instance and it did look potentially spurious, but it is a single instance.

[kkinfy]

@NickMNS I am curiously waiting to see the results of your analysis. For your information (if that helps ): I don't use AMP. I believe many others who posted in this thread may not be using AMP. Hence AMP may not be the cause and could be a coincidence.

As an additional query: Is your traffic entirely from tier 1 countries?

[NickMNS]

There isn't going to be much analysis to come. Baring AMP the rest of my clicks appear legitimate. I know that there are not that many people using AMP, but it remains relevant because AMP like Android-Webview and Safari-inApp browsers is displaying the content in a sand-box like environment. Unfortunately I get very little traffic from social-media so I don't see much Android-Webview.

My traffic is mostly, US,Canada, UK, Australia in exponentially decreasing order. These 4 make up 90% of traffic on most days.

Let's see if other chime in about November 2, 2020.

[kkinfy]

@NickMNS If AMP pages were responsible, I fear that in my case, maybe in-app browsers could be causing the issue. Is there a way in analytics or apache logs to check if the browser was in-app?

Additional Doubt: If AMP pages were creating invalid clicks, will they not skew Adsense data and hence RPM, auctions, and in turn the earnings?

[yaashul]

I am not using AMP but I see almost 3-4 times clicks in analytics (sometime even 9 times).

[kkinfy]

@NickMNS and others: I am going through my analytics data for outliers. In few cases, I can see pages that show 0 publisher impressions but 3 to 5 clicks. What on earth could be happening? How can someone click on an ad that does not exist? Is someone clicking on a cached/downloaded page? Do others see this kind of behaviour (only a few cases).

[NickMNS]

@Yaashul when did this start? Please check you analytics pre-November 2020 and then after the start of November 2020.

@kkinfy this the "bug" or whatever you want to call it. Others have reported the same. It does not reflect reality.

Is there a way in analytics or apache logs to check if the browser was in-app?

In the "publisher pages" report add a secondary "users" and then "browser", this will show the browser for each page. "Android Webview" is one and the other is "Safari (in-app)"

If AMP pages were creating invalid clicks,

I'm not suggesting that AMP is creating the invalid clicks, I actually doubt that there are any invalid clicks at all, what I think is that there is a bug with the "First-Party Cookie" mechanism that is occurring when pages are displayed in anything other than a standard browser.

[yaashul]

For me it started about 2 and half year ago. When I was on ADx and integrated analytics with ADx. Then I shifted back to adsense and it continues.