Tags: analytics commerce concept
View-Through Attribution
Date: 2026-08-16
Crediting an ad nobody clicked. It’s defensible in principle — display and video do work without clicks — and in practice it credits impressions to people who were going to buy anyway, which is why it flatters retargeting most.
What it is
View-through attribution credits a conversion to an ad impression the user saw but didn’t click, within a lookback window.
Click-through requires an action. View-through requires only that the ad was served, and often only that it was served, not that it was seen.
The case for it
Genuine, and worth stating before the criticism:
- Display and video mostly don’t get clicked. Click-through rates are a fraction of a percent; if impressions did nothing, brand advertising wouldn’t work at all
- The click is an arbitrary boundary. Someone who sees an ad and searches your brand later was influenced. Click-through attribution records that as organic or branded search
- It’s how the channel is bought. Ignoring view-through means display and video appear to produce nothing, which pushes budget entirely to bottom-funnel channels
The case against
The mechanism has a specific flaw: an impression is trivially cheap to deliver to people already likely to convert.
retargeting: serve an impression to everyone who
visited a product page in the last 7 days
↓
those people were already likely to buy
↓
a share of them buy
↓
all of those conversions are "view-through"
Nothing in that chain requires the ad to have done anything. It’s Correlation and Causation with a budget attached, and it’s structurally worst for retargeting — the channel where view-through credit is claimed most.
The window compounds it. A one-day view-through window is a defensible claim about recency; a 30-day one credits an impression scrolled past a month ago.
The impression standard
Worth knowing what “viewed” means, because it’s weaker than it sounds. A served impression may never have entered the viewport. Viewability standards typically require some proportion of pixels in view for some duration, and even that doesn’t mean a person looked at it.
So “view-through conversion” can mean “an ad was delivered to a browser that later converted”, with no evidence a human perceived it. [CHECK: current viewability standards and whether a given platform reports served or viewable impressions.]
Reading platform view-through numbers
- Every platform counts its own, on its own data, with its own window and no visibility of the others. Sum them and you’ll exceed your total orders — Walled Garden Reporting
- They rarely appear in your analytics at all, because you never saw the impression. So the platform’s number and yours describe different universes
- Click-through and view-through should always be reported separately. A blended “conversions” figure hides which kind you’re looking at, and platforms default to blending
The only way to settle it
Withhold the ads from a random group and compare. A geo holdout or an audience holdout answers directly whether the impressions caused anything — Incrementality Testing, Geo Holdout Tests.
The consistent finding when brands run this against retargeting is that a substantial share of view-through credit is not incremental. That’s not an argument for zero display spend; it’s an argument for sizing display from a holdout rather than from platform-reported view-through.
A working position
- Report click-through and view-through separately, always
- Short view-through windows — a day, not a month
- Treat view-through as directional, not as revenue
- Size upper-funnel channels with incrementality, and use platform reporting for within-channel optimisation where the bias is constant on both sides
- Never let view-through into a blended return-on-ad-spend figure used for budget allocation. That’s where the over-credit does its damage