If you scrap a website, you'll know how what % of times an ad shows up.

Now, if you know the distribution for top N ads. You can skip everything and go straight to the ad network and check their total number of impression volume on particular website/GEO. You can pretend to be an advertiser and easily get access to this info.

Now, you know the distribution and total, can't you work out the impression count for each ad?