40-50% Less Impressions Than They Are :-(, I have 7 mil. page impressions a month and OpenX ad shows 3 milijons Options

[jerry2]

Here is my problem:

I have a site that has 7 mil. monthly impressions. I count them every time a page is opened and there is no ajax to misguide that track.

I installed OpenX few months ago and I put a banner 160 x 600 on out 160 x 600 position that is on EVERY page as remnant and no other limits or channel limits or what so ever.

It is working great, every time I open any page on my site the banner is shown as should be. But I found out, in OpenX statistics the impressions show 40-50% less that I have open pages. How can this be? I have no Ajax, my counter is immediately after the OpenX javascript and I only update a pointer x=x+1 in my database, that can not be wrong.

Ok, somebody may have ad blocker program to stop ad but not my counter in server side, but 3.5 milijon difference is impoosible that way.

Google analytics and ADAware are both showing around 6-7 milion impressions so the count is there around, certanly not 3.5 milion...

Any ideas? I would be gratefull as having 3.5 milion impressions is a lot worse for advertiser revenue than 7 milions.

Thank you for your time

Jerry

[jerry2]

I am wondering why this kind of questions don't get answered. I can not believe thousands of people are using OpenX and nobody noticed that the impression count is wrong (there are 60% more impressions shown that the statistics says) ;-(

[polapatl]

I am wondering why this kind of questions don't get answered. I can not believe thousands of people are using OpenX and nobody noticed that the impression count is wrong (there are 60% more impressions shown that the statistics says) ;-(

I've found the same problem and I don't know how to answer the media company buying space from me. They also asked about CTR which will also show as very low. I can't answer them about that either. I am looking into another solution such as PHP Click Counter to count the CTR. I stand to lose '000s of $ if I don't get an answer.

[rakshan]

They also asked about CTR which will also show as very low. I can't answer them about that either. I am looking into another solution such as PHP Click Counter to count the CTR. I stand to lose '000s of $ if I don't get an answer.

Shouldnt the CTR go UP if impression count is low?

I think most people are not responding to your thread because we dont see these discrepancies in impression counting.

Have you tried installing openx on a test box and doing a controlled test with different browsers?
Have you tried switching your delivery mechanism (from single js call to individual js calls, from js to iframe etc?)

[jerry2]

They also asked about CTR which will also show as

Have you tried installing openx on a test box and doing a controlled test with different browsers?
Have you tried switching your delivery mechanism (from single js call to individual js calls, from js to iframe etc?)

I have OpenX only on one box, didn't try another machine.

I tried switching from JS to Iframe, I don't have single JS option.

My pages are counted on the server side. Everybody who opens a page right below the javascript for openx I have in ASP/Mysql:

UPDATE counter set impression=impression+1

I think this code can not be wrong. As I have no AJAX on my page everybody who opens a page opens a whole page, there is no refresh of the server side without refreshing the ad.

Some people may have ad blocked (firefox module) I understand this discrepancy, but 50% (3.5 mil. impressions???). I have triple checked that I have the random number inserted as random every time the page refreshes.

I have NOT SEEN yet with any browser the page that is NOT shown with banner. It allways shows.

From all this I am stil certain it is a bug, I am a programmer for 20 years and I have a good "nose" for bugs. It seems it doesn't happen to all users :-(

I do have a Windows 2008 server, OpenX should need Linux, but if everything else works perfectly I would suspect the OS can not be a reason for this problem.

Thank you for your time.

[rakshan]

I have OpenX only on one box, didn't try another machine.

I tried switching from JS to Iframe, I don't have single JS option.

My pages are counted on the server side. Everybody who opens a page right below the javascript for openx I have in ASP/Mysql:

UPDATE counter set impression=impression+1

I think this code can not be wrong. As I have no AJAX on my page everybody who opens a page opens a whole page, there is no refresh of the server side without refreshing the ad.

Some people may have ad blocked (firefox module) I understand this discrepancy, but 50% (3.5 mil. impressions???). I have triple checked that I have the random number inserted as random every time the page refreshes.

I have NOT SEEN yet with any browser the page that is NOT shown with banner. It allways shows.

From all this I am stil certain it is a bug, I am a programmer for 20 years and I have a good "nose" for bugs. It seems it doesn't happen to all users :-(

I do have a Windows 2008 server, OpenX should need Linux, but if everything else works perfectly I would suspect the OS can not be a reason for this problem.

Thank you for your time.

50% is a large discrepancy.
I don't think there is a bug in the code or else everyone would be screaming bloody murder.
A lot of large companies use openx, and if you are running your own adserver, you have full control over the code and you can see that the adserver has no reason to hide impression.
Openx doesn't even filter out fraud impression/clicks (which is a much needed feature).

The OS should not be a reason, but it makes a huge difference (i.e. file paths, permissions, crons etc are all implemented differently). and I'm really not sure if there are enough people running openx on windows server (we use linux), so who knows, maybe it is a bug on windows server.

You've seen this discrepancy from the very beginning?
When you drill down to hourly stats, you still see this?
The maintenance cron isnt reporting any errors in debug.log?
If you open your website in firefox with firebug plugin, you can see calls going out to lg.php and the cache buster param is working as intended ?

A controlled test environment would help because you can then view 10 pages of ads on the test env and expect 10 impressions to be counted.
Then you can also use a service like browsershots.org and see results in different browsers and expect a reasonable impression count.

[jerry2]

You've seen this discrepancy from the very beginning?
When you drill down to hourly stats, you still see this?
The maintenance cron isnt reporting any errors in debug.log?
If you open your website in firefox with firebug plugin, you can see calls going out to lg.php and the cache buster param is working as intended ?

A controlled test environment would help because you can then view 10 pages of ads on the test env and expect 10 impressions to be counted.
Then you can also use a service like browsershots.org and see results in different browsers and expect a reasonable impression count.

Thank you for your time...

Yes this discrepancy is from the very beginning.
Using hourly stats it is still there, I have tested that.
Only bugs in debug.log are the maintenance can not start, probobaly because it start's too often. I don't se this as a problem.
How can I test the ads going to lg.php with firefox? The cb parameter is changing yes.

I understand the controlled environment. I also know that if everybody would have 50% less impressions it would be insane. Must be something specific.

Maybe I try putting Allow PHP code in ads to on (security risk) and then increasing the DB counter in the banner code, I could see more accurate how many times banner was shown.

The banners do load every time and impression is counted even if they are cached? I hope and think this is how it works. Because on every page the banner is not realoaded but the PHP code is executed. I guess the banner is in browser cache but the impression is counted.

Yours

Jerry

[rakshan]

Yes this discrepancy is from the very beginning.
Using hourly stats it is still there, I have tested that.
Only bugs in debug.log are the maintenance can not start, probobaly because it start's too often. I don't se this as a problem.
How can I test the ads going to lg.php with firefox? The cb parameter is changing yes.

I understand the controlled environment. I also know that if everybody would have 50% less impressions it would be insane. Must be something specific.

Maybe I try putting Allow PHP code in ads to on (security risk) and then increasing the DB counter in the banner code, I could see more accurate how many times banner was shown.

The banners do load every time and impression is counted even if they are cached? I hope and think this is how it works. Because on every page the banner is not realoaded but the PHP code is executed. I guess the banner is in browser cache but the impression is counted.

Yours

Jerry

if its been there since the beginning, then you're doing something wrong in use or config of the application.
You should switch to scheduled maintenance from automatic maintenance. Set the cron to run hourly.
You can see the requests going to lg.php by using firebug plugin in firefox.

Impressions will not be counted if you dont bust the cache. You can see it in firebug.
If your html banner code has cache-busting javascript, only then it is ok to allow caching, because the javascript will execute and make fresh requests. In all other cases, cache busting is a must.

If you cant create a controlled environment, create a separate "test" advertiser, add a "test" campaign and a test banner inside it. Attach a "test" zone to the banner and put the invocation code on a plain html page. Then reload the page 10 times using different browsers and see if you see the correct results in impression stats.

Note that you have to wait 20 minutes generally after creating banner/zone to test it.

[jerry2]

I have firebug but I don't know where to look for the lg.php.

Cache busting is cb random code? I do that, but I see javascript already does that also by itself.

I'll try the test you recomended. Thanx.

[rakshan]

I have firebug but I don't know where to look for the lg.php.

Cache busting is cb random code? I do that, but I see javascript already does that also by itself.

I'll try the test you recomended. Thanx.

firebug has a "net" tab which shows all the calls being made. u can look for lg.php there. You can also click on a specific request and see response/params etc.

[qqq]

i have same problem
in stat shown by interfeice for zone_id=274 by days I see
07/07/2011 1,317 4 0.30%
select sum(impressions), sum(clicks) from phpads_data_intermediate_ad where date_time> '2011-07-07' and date_time< '2011-07-08' and zone_id=274;
+------------------+-------------+
| sum(impressions) | sum(clicks) |
+------------------+-------------+
| 6282 | 4 |
+------------------+-------------+

Its appear not for all zone!!!
After some time managers see that great difference in statistics appear, in DB all right, but in UI something strange.