Promotions: what you ran, what your category ran, and what it produced
Last updated: 27 September 2026
The Promotions chapter answers one question: did you discount as much as your category during a given window, and did it produce anything. It reads price, Buy Box holding and sales rank before, during and after, and it separates the operations Amazon announced from the promotional waves it never did.
The chapter opens on the only question a vendor needs answered first: during this window, did you discount more or less than your category, and is that your habit. You funded your deals; the chapter tells you what they produced.
For every product it shows the advised price, the Buy Box price before the window and the lowest price during it, the gap to both, the sales rank before, during and after, whether the price came back up, and who held the Buy Box. A cut while you held the Buy Box is your funded operation. A cut while a third party held it is erosion you did not pay for. The chapter separates the two, and says that this separation is inferred from holding, not read from a badge.
A price cut is read in two ways. The main figure compares the lowest Buy Box price during the window with the product’s usual price, the median of the weeks before: did the event move prices? Next to it, the chapter shows how many of those cuts were also a new low for the month: at least as deep against the lowest Buy Box price of the previous 30 days (EU Omnibus rule), earlier promotions included. That is the reference a discount announcement is measured against in the European Union. The two differ because Buy Box prices move often: many event prices had already been reached a few weeks earlier, so a cut against the usual price is not always a new low. The second figure is shown as a share of the cuts, with its count, and never as a separate verdict.
It works with three kinds of periods. Official windows are the known commercial events, maintained as a calendar with explicit dates per marketplace. Control windows are quiet periods used to measure the base rate: what your listings do in an ordinary week. They follow the rule of the events they are compared with: the same length (a 2-day event is compared with 2-day ordinary windows) and the same reference period before the window. An event with no ordinary window of its length, such as Black Friday or Back to School, is shown as not comparable rather than compared with a shorter one. And detected windows are promotional periods inferred from observed price behaviour.
The base rate is what makes the chapter honest. A price drop or a Buy Box rotation during a big window only means something compared with how often the same thing happens on an ordinary day. Every figure is read against that baseline, and against what your category did over the same days.
Detected windows are observations, never calendar entries: the service reports that prices behaved like a promotion, it does not decide that an official operation took place. The two are kept visibly separate so a detected discount wave is never confused with a planned campaign. And a window that was measured with nothing discounted says so in those words: it is a finding, not a gap in the data.