What we check before trusting a display impression count

A reported impression and a genuinely seen impression are not the same thing, and the gap between the two is bigger than most display campaigns account for. An ad served below the fold, in a background browser tab, or to a bot rather than a real visitor still counts as a delivered impression in a raw report, inflating a number that had no chance of ever reaching an actual person.

Viewability is the first check: whether an ad actually rendered in a viewable position on the page for a meaningful duration, not just whether it was technically served. Industry benchmarks treat an ad as viewable only if at least half of it was on screen for a second or more, and a placement with a low viewability rate is effectively wasting a meaningful share of its budget on impressions nobody had a real chance to see, regardless of how the raw delivery numbers look.

Fraud detection is a separate check, screening for non-human traffic, bot networks inflating impression and click counts on low-quality inventory. This matters more on cheaper, less-vetted placements, where fraud rates tend to run higher precisely because the low cost that made the placement attractive is sometimes a symptom of exactly this problem, not a genuine efficiency.

We only report and optimize against verified impressions, ones that pass both the viewability and fraud checks, rather than raw delivered volume. A placement that looks efficient on cost per impression using raw numbers can look considerably worse once verified impressions are the actual denominator, and that adjusted view is what should drive budget decisions, not the inflated headline number a low-quality placement would otherwise show.

Checking viewability and fraud before trusting a number is not a nice-to-have layered on top of a display campaign, it is what makes the rest of the campaign's optimization decisions actually valid, since reallocating budget based on unverified impression counts just reallocates toward whichever placement is best at generating fake or unseen delivery, which is the opposite of what a programmatic campaign is supposed to be doing with real behavioral data behind it.