counting methods -- need reassurance

am I understanding this correctly?

[arloleach]

Hello,

I'm working with a homegrown ad rotation system, and I'm in the process of upgrading it for compliance against the IAB measurement standards. Specifically, I've just moved from a server-initiated to a client-initiated counting method, so that the counting occurs as late in the page display process as possible.

Out of curiosity, I left both counting methods enabled so that I could compare their results side by side. I expected the client-initiated method to yield a lower result, of course, but I didn't expect it to be this much lower: I'm seeing that the client only receives about 70% of the ads that the server serves up.

Does this seem normal? Can anyone list the typical reasons for such a large discrepancy? I just want to understand this better before I bite the bullet and switch to a counting method that will give me so many fewer impressions than I thought I was getting!

Thanks much,
-Arlo

[arloleach]

In case the IAB's "server-initiated" and "client-initiated" terms weren't clear enough ... we used to log the impression at the time the content management server was requested to serve up a page. Now we wait until the web browser has retrieved the assembled HTML page, requests the particular ad graphic, and the ad graphic has been served to the browser, to log the impression.

[AlienPsychic51]

I’m just shooting at this with an idea. Since you’re seeing a discrepancy between ads served and ads received it could it be due to loss through the network in between. As far as I know while the page is built on the client side requests are made to the server for images to be placed within the page. If an image is corrupted or lost in the transition then another request is made at the network level for another image to replace the one that was lost. The client would only see the exact number of images that it asked for on the application level, but the server would see more.

Actually, client based counting sounds like a good idea. It's probably more accurate, but don't forget that some people might try to tinker with the numbers if they can get access to them.