Attribution Windows on Amazon: Every Surface Compared

Why one number can't describe this system

Amazon runs five meaningfully different measurement systems under the umbrella of "Amazon advertising," and each has its own window, its own eligible interaction type, and — as of January 2026 — a materially different methodology for one of the two crediting types. A single quoted figure for "Amazon's attribution window" is almost always describing one specific surface, correctly, while implying a generality that doesn't hold anywhere else in the system. This page exists as the single reference to check before repeating a window figure anywhere else.

Sponsored Products

Click-attributed only, with no formal view-through mechanism. The click window runs 7 days for sellers and 14 days for vendors. If a shopper sees a Sponsored Products ad without clicking and buys later, that purchase generally isn't credited to the ad through a view-based path the way it can be for other Amazon ad products.

Sponsored Brands and Sponsored Display

Click attribution for both runs a 14-day window, and this specific piece was unaffected by the January 2026 update — Amazon's own update page states click-based attribution remains unchanged across every ad type. View-based crediting is where the change matters: for campaigns billed on a viewable-impression (vCPM) basis specifically, the blanket 14-day view window was replaced by a shopping-signal enhanced last-touch model — shorter, and judged on whether the view plausibly influenced the purchase rather than simply preceding it within the window. Campaigns on other billing structures within these two ad types aren't named as affected by Amazon's own update.

Amazon DSP — the split most explainers get wrong

DSP is the surface where accuracy breaks down most often, because its attribution behaviour genuinely differs by where the ad delivers. Amazon's own update names "Amazon DSP showing in-store ads" as affected by the January 2026 change — DSP inventory serving on Amazon's own retail surfaces moved to the same shorter, machine-judged view model as Sponsored Brands and Sponsored Display vCPM. DSP inventory delivered offsite — the open web, third-party apps, exchanges — is not named in Amazon's update, and by omission continues under the attribution structure DSP has generally used, a click and view lookback most commonly described as 14 days. Any DSP performance comparison spanning before and after January 2026 needs to account for which supply source is being measured.

Amazon Attribution

A genuinely separate system from the four above. Amazon Attribution measures the impact of off-Amazon marketing channels — Google, Meta, TikTok, email, influencer, affiliate — on on-Amazon shopping activity, using its own 14-day last-click model. It's a click-based system with no view-through component, and it's worth not conflating with sponsored ads or DSP windows, since it measures entirely different ad activity: channels that never touch Amazon's own ad platform at all, pointed at an Amazon destination.

The common mistake, including ours

The mistake is treating any single one of these five windows as representative of "Amazon" as a whole, particularly in content written before January 1, 2026 that's never been revisited. We've had to correct our own materials that stated a blanket 14-day view window for Amazon advertising generally — accurate before the update, materially incomplete after it, and exactly the kind of imprecise claim that erodes trust with a reader who happens to check it against Amazon's current documentation.

The broader lesson generalises beyond attribution windows specifically: any Amazon-specific fact that carries a number — a fee percentage, a surcharge threshold, an attribution window — needs the date it was checked attached to it, because Amazon's own documentation genuinely changes underneath a static page. A number without a checked-on date is a number nobody, including the person who wrote it, can vouch for six months later.

Why the DSP split matters more than it looks like it should

The in-store versus offsite distinction inside DSP isn't a minor footnote — it changes what a reported performance shift actually means. An account running a mix of in-store and offsite DSP inventory that shows a view-through conversion drop after January 2026 might be seeing a genuine reporting-methodology effect on the in-store portion and no change at all on the offsite portion, blended together into one confusing combined number. Breaking DSP reporting out by supply source, not just by campaign, is the only way to tell which part of a post-2026 shift is real and which part is an artifact of which placements the new model touched.

How to use this table correctly in your own reporting

Name the specific ad product and, for Sponsored Brands, Sponsored Display and DSP, the billing type and supply source, every time an attribution window appears in a report or a client conversation. When comparing performance across a date range spanning January 1, 2026, flag the methodology change explicitly for any affected surface rather than treating the before-and-after numbers as a clean apples-to-apples comparison — they're measuring meaningfully different things for view-based credit specifically.