Prime Video Buying Read Against a Holdout, Not Attribution

Prime Video shows advertisements to subscribers by default unless they pay for the ad-free upgrade, which is why the reach arrived overnight and why rates settled below what Netflix and Disney were asking. The buying question is not whether to take the inventory. It is what you are entitled to conclude from Amazon Marketing Cloud afterwards, and the answer is less than the deck implies. The clean room is a query environment, not a verdict: The clean room will join your impressions to Amazon's conversion events and return a number. That number is path based by default: it credits an advertisement that appeared before a purchase, inside a store where a large share of those purchases were going to happen regardless. We use it for the questions it answers well — audience overlap, path length, frequency saturation curves, incremental reach against your sponsored placements — and we run the causal question as a holdout, either a suppressed audience inside the same instance or held-out geography where the catalogue permits it. When a client wants the default attributed figure, we supply it, labelled as attributed, printed next to the holdout result. Retail readiness before flight: For anything selling on Amazon we inspect the product pages before booking media, and we will delay a flight over what we find. Buy Box ownership, stock across the marketplaces in scope, review count and rating, enhanced content, and whether the variation family is arranged so the advertisement and the landing page agree. Television sends people to search for the brand; if what they find is a page that has lost the Buy Box, the campaign has paid for somebody else's sale. This is the most frequent reason a Prime Video flight underperforms, and it has nothing to do with the buy. Non-endemic work: Energy retailers and charging networks have no catalogue to sell against, so the conversion signal has to come from your side: the Amazon tag plus a server-side conversion feed, installed and validated before the flight, carrying the events that matter rather than a page view. Where that instrumentation does not exist, the first two weeks of the engagement build it, and we write that into the plan rather than discovering it in week four. Formats we use and formats we skip: Pause advertisements and the interactive units are worth buying when there is something to do with the interaction — a store, a locator, a sample request. Bought as a brand impression with a clickable overlay nobody touches, they cost more than a standard slot for the same delivery. We buy them for a stated reason or we do not buy them. Turkey: Prime Video advertising has not opened in Turkey. Turkish exporters — food and drink brands selling into Germany and the United Kingdom in particular — buy those marketplaces from here, and that is most of our Turkish work in this channel. If the goal is the domestic audience, this is the wrong channel and we will say so on the first call. What we refuse: Percentage-of-media fees, here and everywhere else. Reporting attributed sales without the holdout beside them. Flighting against a catalogue that is not ready, however firm the launch date has been made internally. And running the clean room as a dashboard tour, which produces charts nobody acts on. Who this is not for: Brands wanting a fast test at low spend. The holdout needs scale to resolve and a suppressed cell costs reach, so a small flight will produce a number you cannot defend in a board meeting. What you receive: The campaign live in Amazon DSP, the clean room queried against a written question list, a holdout design with its read date, the retail readiness check naming what we would fix first, and a plain statement of what the flight did and did not establish.