Most products arrive on Amazon the same way. Somebody had an idea, the factory made it, and then, right at the end, someone opens a keyword tool to work out how to describe the thing that already exists.
We ran that process backwards once, and the results changed how we think about launches.
This is a story about Amazon keyword research as product development, not as an SEO chore. We read the search data first, found demand that nobody was serving properly and built a personal care product to fit the searches. It converted at 25% from day one.
Worth telling properly, because the method matters more than the product.
The search bar is a market research department you're not using
Every day, millions of people type exactly what they want into Amazon. Not what they might want, not what they told a survey. What they are actively trying to buy, right now, with a card in their hand.
That data is sitting in Brand Registry's search analytics, in the search term reports of anyone running ads in the category and in the suggestions Amazon autocompletes when you start typing. It tells you the words people use, the problems they're trying to solve and how big each pocket of demand is.
Most brands only ever mine it for phrases to stuff into a title. The more interesting question almost nobody asks: what are people searching for that the current results answer badly?
What the keyword research found, and what we built
Digging through the category's search terms, a pattern kept surfacing. Steady, specific demand, and results that only half-matched it. People searching with clear intent were landing on products aimed at a slightly different job, with reviews full of the phrase "doesn't really work for".
That gap is the whole opportunity. Demand that's already typed out and measurable, being served by compromise. The volumes told us roughly what the market was worth. The reviews on the near-miss products told us exactly what to fix. And the phrasing handed us the listing copy, because we were simply echoing the language of the search back at the searcher.
So the product was specced against the searches. The listing wasn't optimised after the fact. It was the starting document.
Why we only put £150 behind the launch
When the product is the answer to a typed question, launch advertising stops being persuasion and starts being introduction. You bid on the exact searches you built the product for, your listing mirrors the searcher's own words back to them and conversion does the rest.
Ours converted at 25% from launch. For context, a decent Amazon conversion rate is high single digits.
The small ad budget needs explaining, because on its own it reads like either bragging or under-investment, and it was neither. We treated the £150 as a validation budget: a deliberately small test to prove real customers behaved the way the data said they would, before proper money went in. They did. The stock sold through and the ads paid back more than three times what they cost. The real lesson: prove demand on small money and the case for a bigger budget writes itself, faster than you'd expect.
No launch discounting to climb out of later. No months of paying for clicks while the listing "gathered data". The data-gathering had happened before the product existed.
We under-ordered, obviously. Selling out is the good version of a stock problem, but it's still a stock problem, and it's one reason stock depth sits on our launch checklist with a weight that surprises people.
You don't need a new product to use this
The purest version of this method builds the product from the data. But the same research sharpens a launch for a product that already exists, and that's where most brands should start.
When we run this work ahead of a launch, the same things keep happening. Titles come out different. Images come out different, because the data shows which use case to photograph. Sometimes the product itself changes, a bundle here, a size variant there, because demand turns out to sit a millimetre to the left of what was planned.
The test is simple enough to say out loud. Does your planned listing use the customer's words or your factory's words? Answering it properly takes a week of deeply unglamorous digging, which is presumably why so few people do it. That week routinely decides a five-figure launch budget.
The quiet condition that made it work
One more thing, because it's the part that gets left out of tidy launch stories. This worked because we controlled the channel completely. One listing, one seller, our price, our content. Every click landed on a page we'd built, at a price we'd set.
Run the same playbook on a listing where three resellers are undercutting you and the economics fall apart, because the traffic you pay for converts on somebody else's buy box. We've written about why prices slide when you don't control who sells, and launches feel that leak before anything else does. Most brands think they have a marketing problem. Usually they have a control problem, and it starts leaking on day one.
If you're sitting on a launch, or a product that launched flat, the search data almost always has an opinion about why. Hearing what it says costs you nothing.
.png)
.png)
.svg){kind=icon}
