Amazon ranks on sales, not on relevance
This is the thing that makes Amazon SEO different from every other kind, and almost every mistake follows from missing it. Google is trying to answer a question. Amazon is trying to complete a sale — so it ranks a listing on the probability that a search ends in a purchase. Relevance gets you considered; conversion history on that exact term is what actually moves you.
The consequence catches people out: ranking for a term you convert badly on makes your position worse, not better. Impressions with no orders tell Amazon the match was wrong, and the term decays. You cannot force your way onto a head keyword you have no business on — and if you buy your way there with ads, you are paying to teach the algorithm that your listing does not deserve it.
It also means the listing and the campaign are one system. A keyword strategy handed to a listing that does not convert produces expensive traffic and a worse organic position at the same time. That is why we will always push to fix the page before spending on ads for it.
Where the data actually comes from
The most reliable keyword data for your catalogue is not in any third-party tool — it is in Seller Central, it is free, and it is specific to your ASINs. Search Query Performance shows the real queries that produced impressions, clicks and purchases on your listings, with your share of each. The search-terms report from Sponsored Products shows every literal phrase a shopper typed that triggered your ad, with spend and orders against it.
Two weeks of a modest auto campaign buys real conversion data on your own products. A tool subscription buys an estimate of it. We use both, but in that order — and when they disagree, your own data wins.
We wrote the full method up in our keyword research guide, including the filter that takes a 400-row export down to the 30 terms actually worth writing to.
What we do not promise
No percentage. Not because results are not achievable, but because a number quoted without your margin, your category, your review count and your starting position is a number invented to close a meeting. Ask any agency quoting one what the baseline was, and watch what happens.
What we will tell you before you commit: whether your listing is the constraint or your traffic is, what we would change first, and roughly what it costs. If the honest answer is that your images are the problem and you do not need a growth retainer yet, that is the answer you get — it is a smaller invoice now and a better one later.
ACoS is judged the same way. There is no universal good number; a launch campaign at an alarming ACoS can be correct if it is buying rank you will keep. Campaigns get judged on total profit, with the organic share of revenue tracked alongside, because ads that quietly replace organic sales look like growth on the ad dashboard and are not.