difference between server log page views and Google impressions

[CPCretirement]

How big of a difference are most of you seeing between actual page views by server stats and Adsense stats on pageviews? I see a big difference.

For example, I served just over 1,000,000 page views on 9/11 according to my server stats. Adsense shows just under 300,000 impressions. Is this normal?

It doesn't really matter much since the bottomline is how much I get paid. It just seems to be way off.

[Sharper]

Is that including or eliminating bots like Googlebot, Mediapartner, slurp, etc...? Are you only counting pages with adsense on them as page views? (In other words, do you have pages that have multiple html sections, pulled from different files?)

Do you have an audience heavy with older browser types? (Netscape 4.0?)

I find my impressions to be generally about 10-15% lower that logged page views from real people, which is about what I'd expect.

[justageek]

That seems to be a large difference but it is entirely possible. Remember that visitors must be able to handle <iframes> and java script for AdSense to work properly. So page view versus impressions will always be different.

JAG

[CPCretirement]

Adsense is on almost every page. It's certainly on the high volume ones. I am not eliminating bots in those stats.

The discussion forums count for the largest percentage of those views. I wonder if there is some glitch with the way it handles new pages. The members are constantly creating new threads. It can take a few views on each thread before targetted ads show up.

What if someone looks at the same page many times? For example, the topic lists are viewed multiple times each session by users as they browse for threads to read.

[richmondsteve]

I wonder if there is some glitch with the way it handles new pages. The members are constantly creating new threads. It can take a few views on each thread before targetted ads show up.

I haven't seen a glitch. And PSAs that are displayed count as impressions. Clicks on them don't count as clicks since they don't load a Google page, but they do count as impressions.

What if someone looks at the same page many times? For example, the topic lists are viewed multiple times each session by users as they browse for threads to read.

It counts as an impression. If the JavaScript executes it counts as an impression. If it doesn't execute, no impression. Quite simple - page views that don't have AdSense code enabled, views from bots, link verifiers, etc. and views from text browsers like lynx, browsers without IFRAME support and browsers with JavaScript disabled will not generate AdSense impressions.

[MarkJH]

I didn't think that PSA's were counted as impressions? That would certainly explain the difference between the stats.

[justageek]

I think most of the things that would cause the stats to be different have been identified. The only other thing that might do it is that you mentioned the ads were on a discussion forum and it takes a few page views before the targeted ads show up. If that's the case then that means a Google spider is most likely camped out on those pages which would show in your log and not as an impression. So, if you mean few views as in maybe 4 or 5 then about 20+/- percent of your hits will be the camped out spider.

JAG

[CPCretirement]

Overall page views have not gone up much since adding the adsense code. I don't think that it's a case of the spider accounting for a large chunk of the traffic.

Strange indeed. Maybe the PSAs aren't counted in stats after all. That would explain a big part of the difference.

Oh well. No big deal either way. I'm happy with the overall $$. That's what really counts. :)

[justageek]

Yes. Strange. Maybe your visitor rate has fallen but the spider is keeping your numbers up. There must be a spider to get the ads correct unless I'm missing something. You may want to check. But then again, as you said, if the revenue is there then that's what counts :-)

[Sharky]

My stats show a ratio of about 2:3 for AdSense impressions vs. Page Views, even after accounting for visits from spiders. As far as I can tell, I'm getting close to zero visits from old browsers. I'm counting page views through a special gif that's on each page with the Google <iframe> code, not through html hits.

I suspect that the problem is related to caching. The number of cached page views that show in my stats is just slightly less than the number of impressions counted by Google. The counter gif that I use is uncached, and its numbers are 50% higher. I haven't run a formal caching test on the Google javascript URL, but I do know that if I browse off of one of my pages and then back onto it, then I see the same ads that were there the first time around.

[morpheus83]

I have adsense running on every page of my site. However between the server log and adsense there is a difference of 10-15% always. I dont understand why google regsiters less impressions.

[Blue_Fin]

See richmondsteve's post above and this is also addressed in the AdSense FAQs.

[communitynews]

Our reports from webtrends were showing page views as 8 times the number of exposures reported on the google adsense admin reports. The first thing we discoved was that webtrends was counting any URL with a "?" as a page view. Because we control caching of GIF files by adding the file modification date to the name like so "filename.gif?20030101" they were being counted as page views. The figure reported by webtrends as document views doesn't count the "?" so we tried to match document views to google adsense exposures and document views were still 4 times more than exposures. We also added a counter to the server every time the google code was served and confirmed these numbers.

For a long time we suspected caching but finally we loaded a days worth of transactions into an access DB and started deleting out images and other files that don't have ads, response codes other than 200, and transactions with zero bytes returned. We then compiled a list of User Agents in the file and removed all from spiders and those browers that don't support IFRAMES.

The remaining transactions matched google adsense within 5%, which we believe represents mostly people with Java disabled.

Bottom line is our problem was the "?" counting by webtrends and the fact that we have far more spider activity than people searching. We were a little shocked that we have so much spider activity but then again we get 80% of our real human traffic from search engines.

[CPCretirement]

Interesting. Our forum software uses a "?" on individual sections and threads. It doesn't use one on the main forum page.

I don't think that we have any overcounting from the? as you do with the images. The files/pages of ours that have the? are actual pages, not just images.

I wonder if Google isn't counting the pages with the question mark in the url. That might account for some of the difference.

On the other hand, maybe something is wrong with my server stats. I might not be getting as much traffic as I think! I know they aren't completely wrong though since my ad serving software also shows a higher number than Google does and it's not on all of the pages that Google is on. But it has always been lower than the page views I see in the stats. Hmmm.... I'm going to have to look into that possibility some more.

[irock]

MarkJH,

Official answer from Google confirms that PSAs count as valid impressions.

[MarkJH]

Thanks, irock.

That's interesting. That means that my AdSense impressions aren't even half what my logs show. Can't be that many people with Javascript switched off!

[irock]

Unless that majority of your audience REALLY hates interactivity, most of them will probably not consider turning off javascript. On my site there's only 2% of whose browsers aren't javascript enabled.

So, I think we should ask Google why the HECK page impressions go down like this.

[communitynews]

CPCretirement, note that counting "?" in a URL that is a page (.html or CGI) and does have google java would not be a problem. Our problem was that we used "?" in displaying images which can't contain the google java.

[communitynews]

MarkJH, you're right that half your page views probably aren't from people with java turned off but my earlier post shows that half your page views could easily be from spiders.

[MarkJH]

Googlebot itself only accounted for 1.7% of all my traffic for the month of September. I find it hard to believe that spiders account for 60-70% of my total traffic.