View-Through Conversions: What They Prove and What They Do Not

What the metric actually measures

View-through conversion counts a sale when a shopper viewed — but didn't click — an ad, and then purchased within the eligible window. Amazon requires the view to meet Media Rating Council viewability standards: at least 50% of pixels in view for a minimum of one second for display ads and two seconds for video. That's a real, meaningful bar for what counts as "seen" — a banner that scrolled past off-screen doesn't qualify — but meeting the viewability bar still only proves the ad was genuinely visible, not that it changed anything about the shopper's decision.

It exists because a large share of upper-funnel ad exposure, by design, doesn't produce a click at all. Video and display ads are built to be seen, not tapped, and a channel judged purely on click-based attribution will look artificially weak simply because clicking was never the expected interaction. View-through attribution is the platform's way of giving that exposure some credit — the question this page is about is how much confidence that credit deserves.

The January 2026 change, and why Amazon made it

Effective January 1, 2026, Amazon replaced the blanket 14-day view-through window with what it calls a shopping-signal enhanced last-touch model, for Sponsored Brands and Sponsored Display campaigns billed on a viewable-impression (vCPM) basis, and for Amazon DSP's in-store placements. The new model evaluates whether a given ad view actually influenced the purchase, using Amazon's own shopper behaviour signals, rather than crediting anyone who saw an ad and bought within two weeks regardless of relevance. Click-based attribution is explicitly unchanged. Advertisers can still access the old-style figure — "purchases, all views" — as a separate 14-day metric inside unified reporting, but it's no longer the headline number.

Worth being precise about scope: Amazon's own update page names the change as applying to those specific vCPM campaign types and DSP's in-store placements. It doesn't name offsite DSP inventory — display and video served on third-party sites and apps — as affected, so that inventory continues to be measured under the standard click and view structure most DSP reporting already uses.

Why the old blind 14-day window was worth fixing

A blanket 14-day view window credits every purchase within that window to the ad, regardless of whether the shopper would have bought anyway. A regular Amazon customer who reorders a household staple every two weeks, and happens to see a display ad for it during that window purely because they browse Amazon often, would count as a view-through conversion even though the ad changed nothing about a purchase that was already going to happen on schedule. That's the specific failure mode the shopping-signal enhanced model is designed to reduce — crediting views that plausibly influenced discovery, rather than every view that merely preceded a purchase.

What view-through conversions still can't prove, even improved

Even a smarter, ML-judged view-attribution model is still fundamentally answering "did this view precede a purchase in a way our model considers plausible," not "would this purchase not have happened without the ad." That second, causal question is what a holdout test or geo-lift study answers, and no attribution model — however sophisticated its judgment of relevance — replaces that test. Treat an improved view-through number as a better-filtered signal, not a proof of incrementality.

The common mistake, including ours

The mistake is reporting a view-through conversion count as if it were equivalent to a click-based sale in strength of evidence, when the two answer different-confidence questions. We've built client summaries in the past that blended click-attributed and view-through-attributed purchases into a single "attributed sales" figure without separating the two, which flattered accounts running heavy view-based DSP or Sponsored Brands vCPM inventory. We now report the two separately, and flag view-through figures with the specific window and model version behind them, particularly given how recently the underlying methodology changed.

How this connects to what runs off-Amazon

The distinction matters most for exactly the inventory where a click is rarest by design — streaming and connected TV. In our own DSP practice, we set streaming a return target near 1.0 and judge it on video completion rate and branded-search lift rather than attributed ROAS alone, because attribution structurally undercounts it: shoppable features generally only fire for logged-in Prime members, so a meaningful share of streaming-driven demand arrives at Amazon unattributed altogether, view-through or otherwise. A tighter, more relevance-aware view-attribution model helps at the margin, but it doesn't solve the deeper measurement gap for inventory this far upper-funnel — which is why we treat streaming's attributed numbers as a partial read, never the full picture.

What to do when your view-through numbers shift after January 2026

If your Sponsored Brands, Sponsored Display or DSP in-store view-through conversions dropped noticeably after the January 2026 change, that's expected — the new model is deliberately more conservative about which views it credits, and a drop in the headline number doesn't necessarily mean the underlying ad exposure changed at all. Before reacting, pull the "all views" 14-day metric alongside the new shopping-signal-enhanced figure for the same period, to separate a genuine performance change from a reporting-methodology change. If the gap between the two is large, that gap itself is informative about how much of your prior view-through reporting was crediting views that likely weren't influencing anything.