What is Attribution Window?

The attribution window is the period after a shopper clicks your ad during which a purchase is still credited back to that click. It is the single most misread mechanic in Amazon reporting, because sales are booked against the day of the click, not the day of the purchase.

Windows by ad type

These are the standard reporting windows, the ones the console shows by default.

Sponsored Products reporting also exposes shorter and longer windows as separate columns, so the column you sum is a deliberate choice rather than whichever one appears first. Mixing ad types is a real reporting bug rather than a rounding issue: a query that sums only the 7 day sales column returns nothing at all for Sponsored Brands and Sponsored Display rows, so totals have to be summed per ad type against that type's own window.

Sales are credited backwards

A hypothetical worked example

The dates below are invented. Suppose a shopper clicks a Sponsored Products ad on the 1st and buys on the 5th. The sale is written against the 1st. If you pulled a report on the 2nd, the 1st looked like spend with no sales. Pull the same report on the 9th and the 1st may be profitable.

The practical consequences:

• Yesterday always looks worse than it will end up. Recent days carry mature spend against immature sales, so ACoS on the trailing few days is inflated by construction.
• A report is not final when you first read it. Rows keep restating until every window that touches them has closed.
• Judging a bid change too early rewards noise. The data that would justify reversing it has not arrived yet.

How we handle it

Our bidding engine reads a 30 day metrics window that ends two days ago, and sums sales per ad type on that type's own window. The doctrine behind it is to weight unsettled attribution rather than skip it, because dropping immature days throws away the freshest signal you have.

Clicks restate too

Attribution governs sales and orders. Clicks and spend have a separate lag: Amazon delivers traffic data before invalid traffic has been filtered out, so clicks and spend can be revised down in the days after they are first reported. Very fresh traffic figures therefore run gross, which matters whenever you compare a live hourly feed against a settled report.