Myth: view-through conversions prove your display and video ads are working

Myth: view-through conversions prove your display and video ads are working

The question: a view-through conversion (VTC) credits an impression a user saw but never clicked, then later converted — doesn't that capture the brand-building value clicks miss?

What the experiments show: VTCs are the single most over-stated number in attribution, because they are vulnerable to selection bias on a massive scale. Programmatic targeting deliberately serves impressions to users most likely to convert. So the people who 'see' your ad are precisely those already predisposed to buy — and crediting their conversions to the impression confuses targeting with persuasion.

The nuance: ghost-ad and PSA-control experiments — where a randomized holdout is served a placebo ad and their conversion rate is compared to the exposed group — repeatedly find that a large share, often the majority, of view-through credit is non-incremental. A well-known class of studies on retargeting found measured ROI collapsed once a proper control isolated the counterfactual. The impression was correlated with conversion; it rarely caused it.

This is selection bias, not measurement error — no attribution window length fixes it, because the bias is baked into who gets the impression.

Bottom line for practitioners: treat view-through numbers as an upper bound contaminated by targeting, never as proof of effect. Validate display and video with a randomized holdout (geo or user-level PSA control). If conversions barely move when you suppress the ads, the view-through credit was an artifact.