observations on adsense "page impressions" vs web server page logs - Google AdSense - Display Ads forum at WebmasterWorld - WebmasterWorld

Forum Moderators: martinibuster

Message Too Old, No Replies

observations on adsense "page impressions" vs web server page logs

nyc863

1:03 am on Jul 10, 2015 (gmt 0)

10+ Year Member

I have been carefully auditing the number of mobile pages and desktop pages delivered by the web server, removing bots which are a ridiculous 2x to 3x the traffic of real users. We are quite careful to limit bots, the ones we let in, such as google, baidu, bing, msn, yahoo etc we can easily remove from stats. Bing for some reason is the #1 bot with twice the page views of googlebot, and the rest are pretty insignificant. 90% of bot traffic is google plus bing plus baidu.

Ok so what we're left with is mobile pages and desktop pages - note I strip out all ajax hits or other non-page hits, and am only counting pages that had 2 to 3 adsense ad units inserted in them, so I'm not including pages that don't have ads for whatever reason.

As a day progresses, adsense 'today' panel agrees on the number of mobile pages to within a few percent. For example today, so far, we have 40k mobile pages and adsense says 39k mobile pages.

However desktop pages are being under-counted by 50% .. for example today so far, counting from 3am east coast time (midnight google rollover time), we have logged 250k desktop pages but adsense reports 110k desktop page impressions.

This has been pretty consistent under-counting it isn't just a temporary thing.

There are three possible explanations, all somewhat unbelievable
1, bots are masquerading as users (but not as mobile users) avoiding all traps and excess reader limits and mostly using US IP addresses and modern browser user agents. They are doing so over a very wide range of IPs as well so no one IP is generating more requests than a power user.

2, People are clicking so fast that the google ad code is not getting the chance to get going (and yet this is not happening for mobile)

3, 50% of desktop visitors are using noscript, or an adblocker, or both noscript and an adblocker.

Thoughts? experiences with adsense "page impression" accounting when you slot in a different ad network?

netmeg

2:18 pm on Jul 10, 2015 (gmt 0)

WebmasterWorld Senior Member

10+ Year Member

Top Contributors Of The Month

Having had extensive experience with good, bad, and ugly bots, my vote would be mostly #1 and possibly a little bit of #3, depending on your niche (i.e. tech users are probably more prone to using an ad blocker than some other civilians would be)

During a couple of my heaviest bot attacks, the bots were pretty likely infected Windows machines, coming in from all over the world - with one attack the browser was reported as assorted versions of IE, and with another it was a specific version of Firefox. One attack visited only one or two pages on the site, and another attack spread it around more. The bots have become very sophisticated, and there is little to discern them from actual users to the naked eyeball.

Now, how Google identifies them as bots, I don't know. They have all kinds of technologies they have hired and/or purchased to combat click fraud in recent years. It has also been speculated that in some cases they could be overcompensating in order to protect the advertisers. Of course, we will never know whether or not that's true.

There's also this help page from Google about AdSense page impressions vs Analytics page impressions:

[support.google.com...]

nyc863

6:29 am on Jul 12, 2015 (gmt 0)

10+ Year Member

By placing an event with each analytics page event I find I am losing 15% right there: 300k analytics impressions and an event on 10% of them, I get 25k events (250k).

the other 50k never fire the event.

but 250k do run full analytics and have time to fire the additional event yet for desktop users, only 50% of those page impressions get counted in adsense.

i have trouble believing the bad bot with full javascript theory. I should add an img beacon after a google iframe has been verified with non zero height,

but if google are going to refuse to count a bunch less page impressions that are counted by analytics it would be nice to know why. Mainly i want to know if the users are clicking through so fast that the ads dont even get a chance to appear.

webcentric

12:35 pm on Jul 12, 2015 (gmt 0)

WebmasterWorld Senior Member

10+ Year Member

Top Contributors Of The Month

I believe that page impressions also require the ad being visible in the viewport e.g. if the ad is below the fold and the user never scrolls down, then I'm pretty sure that's not an impression. Others can probably provide better insight on that topic though.

netmeg

3:35 pm on Jul 12, 2015 (gmt 0)

WebmasterWorld Senior Member

10+ Year Member

Top Contributors Of The Month

I believe that's probably true. One of the 'alerts" that Google sent me was that one of my sidebar ads, which is absolutely above the fold on a desktop, was receiving impressions only 20% of the time. That site is responsive, and that ad unit drops underneath the content on a tablet or mobile device. As it happens, I checked my realtime Analytics stats at one point last week, and they told me 6500 simultaneous users were on the site, and 98% of them were mobile. That's an all time high percentage even for me.

All that leads me to believe that 1) Google doesn't count ad impressions unless the ad unit fully loads on whatever device, and 2) I need to rethink my placements for mobile for next year. It's tricky cause you can't put anything big above the content on a smart phone. But I definitely need to think about it.