Post-click analysis

Some fraud cannot be seen one visit at a time. A single session from a click farm or a careful bot looks ordinary. The pattern only emerges across thousands: the same devices returning, clicks bunched at the same second past the minute, a source whose visitors never spend more than two seconds with the advertiser, or conversions that never arrive.

Post-click analysis looks at this history. It compares sources, placements and sub IDs against each other and against normal human baselines, using time on site, depth of visit, repeat rates and, where available, advertiser conversion data.

Search engines do this at very large scale and it is the reason revenue is first reported as estimated revenue and later corrected to finalised revenue. Clicks judged invalid after the fact are credited back to advertisers, and the publisher sees a deduction or clawback.

An arbitrageur running its own analysis gains time. Spotting a bad placement after two days instead of six weeks limits the loss and shows the feed provider that the partner manages quality actively. It complements pre-click filtering, which stops the obvious cases at the door.

Think of it like this

It is the auditor who reviews a month of till receipts. No single sale looked odd, but together they show one cashier refunding the same item every evening.