Myth: Set a 70% viewability target and you'll get viewed ads

Myth: Set a 70% viewability target and you'll get viewed ads

The standard hygiene advice is to apply a pre-bid viewability filter — buy only inventory predicted to be 70%+ in view — and assume your delivered viewability will land there.

The gap between prediction and delivery is mechanical.

1. Pre-bid viewability is a prediction from a measurement vendor, scored on the bid request before the impression renders. It's a probability, not a guarantee.

2. These scores are typically modeled at the placement or domain level on historical averages. The specific impression you win can deviate sharply from its placement's mean.

3. Selection bias compounds it: every buyer applying the same 70% filter competes for the same predicted-viewable inventory, raising its clearing price. You pay a premium for a prediction, then measure the realized rate post-bid and find it drifted.

4. There's also a measurement-coverage hole — impressions that can't be measured at all (unmeasurable) get excluded from the denominator, flattering the reported number versus reality.

The disciplined approach is to treat the pre-bid score as one input, then reconcile against post-bid measured viewability per placement and feed that delta back into bidding.

Why it matters: a viewability target filters on a forecast, not an outcome. Without closing the loop on measured delivery, you're buying a vendor's confidence interval and calling it a result.