There was always a 30 percent difference in reported page impressions in my case for years. But over the last week it has been nearer 50 percent! I suspect that google are/were having some problems, and ads were not displaying at times or some regions?
Today though SEEMS so far to be back to normal! But I said that yesterday as well...
Not a lot! (well... dunno about JS)
Looking around, there does seem to be something of a fuss going on.
Trust me to start doing my testing at the wrong time!
But it does seem an awfully large discrepancy.
Do you use frames or IFrames?
For each Frame or Iframe within a page, many statistics packages will tally an impression for every frame in a page. 3 Frames on a page will indicate 4 Impressions, Google will only tally one hit.
Your impression count (if correct) should only be about 20 to 30% higher than Googles. (A rough estimate). Google is not going to tally a "refresh" or multiple impressions of the same page by one user (IP address).
If your pages are extremely complicated and slow to load (lots of images) the Google Adsense code may never execute.
Anyone who runs multiple ads on the same page will typically see fewer Ad unit impressions for ad units further down the page. People are leaving the page before all the ad units were served.
Anyway that's some notes on impressions.
|I see by my server logs that a page has been served 1000 times in one day, yet Google tell me there were less than 100 ad impressions |
It's not just Google that exhibits this discrepancy. We encounter it with other advertising agencies. One non-commercial organisation we deal with offers a possible (non-cynical) explanation as to why their system failed to serve the number of items our server recorded; "Our anti cheat system has probably stopped most of them."
In any event, as a publisher I am disatissfied with the large discrepancy between page impressions we serve to non-robots and the advert impressions for which we get paid.
Well, the page(s) in question are very simple. It's an info. site ('how to make a banjo') no scripting, not much decoration.
No frames, but there are several quite large pics ~10kb on some pages - lots of text.
The server logs are quite clean...
I realy think there is a serious discrepancy here but I can't think of any way of testing it out.
Not much point in complaining I suppose. Just be grateful for what I get. ;-)
Scruffy, I run a similar informational site. And I DON'T see such a large discrepancy. I haven't analyzed it page by page, but my stats match up pretty closely with Google's, and when my page impressions drop on the weekend, so do Google's reported impressions.
Is it possible that your site is hosted on a server that is in a Web backwater, and requests from your pages for ads are just not traveling fast enough for the ads to be served? Sorry, that's an imprecise description, but I hope you understand....
Sorry but if this is your best estimate of spider traffic, then you probably don't have a handle on it and it's much higher than you think.
Like hunderdown, I find my AdSense stats to be extremely accurate.
You need to take a hard look at your raw logs, not just for the browser ID but also for spiders that are impersonating human surfers. They sometimes tip their hand by:
(a) requesting many pages per second, far faster than a person would
(b) requesting lots of HTML pages but no images
(c) requesting the robots.txt file (even if they don't intend to obey it)
(d) not populating the referring page field (although some surfers are set up this way as well)
Well, I was basing my estimate on the number of requests for robots.txt but I'll have to sit down and go through the logs more carefully.
I would like to write a logger script that will also log clicks, not only impressions.
Anyone knows if it is possible?
I would have to write a script that 'called home' to my server and updated a counter. Yes, I can do that.
Ah yes, I see - there's nothing new under the sun today.
|I see by my server logs that a page has been served 1000 times in one day... |
You are most likely looking at hits. A hit is when an image is downloaded. A hit is when a CSS file is downloaded. A hit is when a spacer gif is downloaded. If you are reporting 1000 hits but only 100 impressions, then the difference is likely that the 1000 hits are roughly ten elements that download everytime someone visits your page.
That is what accounts for the major discrepancies.
I agree that visits from bots may account for part of the difference as well. For one of my sites, 13% of my visitors (and a similar percentage of my bandwidth) is attributed to bots.
I experience the same. Google analytics shows more than double the impressions google adsense shows.
I noticed this as well when I sometimes, when I open one of my pages, no ad shows up.
|You are most likely looking at hits |
No. Real pages.
scruffy, if you have an ad server like phpadsnew, it'll give you accurate numbers on how many times the ad block is called up... whether or not google serves an ad there is another question, and i have not yet explored the bot situation wrt the ad server.
sometimes you will see bots running amok in your log files... yahoo has been on a major rampage lately, and linksmanager.com is the worst bot that i've ever seen... it sucks up hundreds of megs of data.
image searches can also mess up the ads, because some search engines will no longer display the ads when they frame the page with the photo in it.
I've had one unusual thing occur quite a few times now. It appears to be a person browsing, all the content of a given page is fetched, images, ads, etc., then for no apparent reason the browser GETS the page itself over and over again apparently until the person moves onto a new page.
I had one hit from a broswer like this with a referral from a Google search. The search keywords showed up as number one in my awstats for the month, but it was only one visitor doing one search. The broswer just kept doing the same GET over and over again after the normal page load sequence. This results in hundreds and hundreds of page impressions which I'm sure Google will not count, they're all from the same IP in the same time frame, and probably time zone.
Anyway another farfetched possibility. I haven't figured out what browser or ISP configuration is making this happen yet.
You mean page views?