Attribution for Sellers and Vendors — what counts, what doesn't, and the windows.
Sponsored Ads attribution looks the same for Sellers and Vendors and behaves differently underneath. The attribution windows, the multi-touch logic, and the off-Amazon traffic case that needs Amazon Attribution proper.

Closing Module 3 with attribution: how Amazon decides which orders to credit to which campaign, and where Sellers and Vendors see the same numbers reported differently. Attribution is the one PPC topic where the right defaults depend on which side of the marketplace you sit.
The Sponsored Ads attribution window
Sponsored Ads attributes an order to a campaign if the shopper clicked the ad and then placed the order within 7 days for Sponsored Products and Sponsored Brands, or within 14 days for Sponsored Display. Multiple ad clicks within the window can all contribute — the last-click campaign typically gets the credit on SP / SB; Sponsored Display uses both view-through and click-through attribution with different windows.
This window is fixed and cannot be configured. ACOS and ROAS numbers in the console are calculated against this window. Orders outside the window do not count toward the campaign even if the shopper saw the ad — they show up in organic.
What this means for ACOS interpretation
A campaign with an ACOS of 30% may be driving substantial organic uplift outside the attribution window that the campaign report does not see. This is why TACOS — total advertising cost over total revenue — matters as much as ACOS. A campaign with rising ACOS but falling TACOS is often paying for itself in lifted organic; the campaign report alone will not tell you that.
Seller vs Vendor reporting
Sellers see attribution in the Advertising Console exactly as described. Vendors see the same attribution but with two additional wrinkles: (1) some of the "attributed sales" in Vendor reporting are wholesale POs to Amazon rather than shopper purchases, which makes ACOS noisier on a week-to-week basis; (2) Vendor Central's sales reporting and the Advertising Console's attributed sales sometimes disagree because they pull from different data sources at different cadences.
The practical guidance: Vendors should treat campaign-level ACOS as a directional metric and look at trends over 14–28 day windows rather than acting on weekly noise.
Amazon Attribution — the off-Amazon case
Amazon Attribution is a separate product, available to Brand-Registered sellers, that lets you attribute off-Amazon traffic (Google, Meta, email, influencer) to Amazon conversions. It uses its own click tags and its own attribution windows. The Sponsored Ads console knows nothing about it; the two reporting layers run in parallel.
For brands running meaningful external traffic into Amazon, Amazon Attribution is the only way to see whether that spend converted. Without it, the Amazon side of the funnel is invisible to your external attribution stack.
Closing Module 3
That is the module. Twenty-four episodes covering Sponsored Ads end-to-end: formats and targeting, the bid arithmetic, the auction, the limits, naming and structure, portfolios, dates and budgets, the three bid-strategy families, match types, auto targeting, relevance, keyword loading, per-keyword bids, product and category targeting, Sponsored Display audiences, budget rules, the structural blueprint, and the attribution model that sits underneath every number you read off the console.
The three modules together — Promotions, PPC Fundamentals, PPC Setup & Execution — are the AMALYZE Amazon Marketing course end-to-end. From here, the daily work is iteration: tighter bids, cleaner negatives, sharper category targets, and the structural changes the data tells you to make.
Watch Episode 24: Attribution for Sellers and Vendors (German)
The German walkthrough — attribution for Sellers and Vendors and the difference with Amazon Attribution.
Trust the numbers your dashboards report.
AMALYZE normalises Seller and Vendor attribution into a single dashboard so you can compare campaigns across the line without re-learning the model every time.