The ad rendered in the feed, the user scrolled past without clicking, and a purchase landed on the storefront some hours or days later. If the platform can match that converting user back to the impression inside its view-through window, the conversion is credited to the ad as a view-through. A click-through conversion sits one mechanic over: the user clicked, then converted inside a longer click window, which is the credit assignment most operators picture when they think of a platform-reported sale. The view-through window is a parameter of an attribution model; see attribution window for the broader concept.
How view-through credit is assigned
The mechanic is an identity match. The platform retains the impression event, keyed on whatever identifier it has for the viewer: a logged-in account, a hashed match key delivered via Conversions API, or a pixel cookie on the browser side. When a conversion event arrives carrying a matchable identifier, the platform checks the impression history inside the view-through window and credits the impression if it finds one. Window defaults are short and configurable. Meta and TikTok default to one day on view; Google Display, YouTube, and DV360 carry longer view-windows, commonly up to thirty days when enabled. The reporting default is its own lever: Meta includes view-through in the headline ROAS figure in Ads Manager, while Google’s Conversions column for Search often excludes it, so an apparent gap between two platforms is sometimes a reporting choice rather than a behavioral one.
Why view-through credit has shrunk under iOS 14.5
The identity-matching surface narrowed with the post-2021 privacy stack. iOS ATT cuts the cross-app identifier for users who decline tracking, and Safari’s ITP and Firefox’s ETP block third-party cookies by default — Chrome remains the outlier after Google reversed its planned cookie deprecation in 2025, but ATT, Safari, and Firefox have dropped the share of impressions a platform can deterministically match. CAPI keeps the server-side payload alive for consented users with hashed identifiers, but does not recover the un-consented share. What the platform cannot match is backfilled with modeled conversions: observed and modeled view-through roll into one figure with no operator-facing breakdown. iOS-heavy audiences show a higher modeled share than Android-heavy ones.
How operators should read the gap
View-through is the single biggest mechanical reason platform-reported revenue diverges from storefront-reported revenue. A campaign Meta reports drove 500 purchases at 4x ROAS may show as 200 purchases at 1.6x ROAS in Shopify session data; the missing 300 are view-throughs the storefront analytics never saw as Meta traffic. The gap is structural, not a tracking bug. The forced question is whether those view-through conversions correspond to incremental sales or to ads served to people who would have purchased anyway — what incrementality testing is designed to answer, and the reason MER and blended CAC have gained weight as headline reads. The common operator posture is to treat view-through-inclusive ROAS as an in-platform optimization signal for the bidder and go to MER or lift output when the question is budget allocation across channels.