The same product wins 1 keyword or 91, depending on which ASIN you paste

A folding step stool sells a couple of thousand units a month, holds first place in its subcategory, and carries more than eighty thousand reviews. Paste its ASIN into a keyword tool and it wins one search term.

Paste the parent instead and the same product wins ninety-one.

Neither number is wrong. They are answers to different questions, and almost nothing in the research tooling tells you which question was asked.

Amazon reports the click, not the product

For every search term it ranks, Amazon publishes the three products shoppers clicked most, with each one's share of that term's clicks and of its purchases. That is the whole of it: three names per phrase, and everybody else is absent.

The name Amazon publishes is the ASIN that was clicked. Not the parent, not the product line, the exact variation. A stool that comes in white, teal and grey arrives in the data as three separate products, and shoppers who clicked the white one taught Amazon nothing about the teal one.

This sounds like a technicality. It decides what your research says.

Two thirds of the "one keyword" products are colours

Across the United States in July 2026, Amazon named 1,151,211 products against 1,221,205 search terms. The shape of it, per ASIN:

That first row looks like a statement about products. It is not. Two thirds of the ASINs that win a single search term are variations of a larger product. One colour, one size, sitting under a parent that may be winning a great deal.

The table describes ASINs, exactly as Amazon reports them. Turning it into a statement about products means knowing which ASINs belong together, and that is the step most tools skip without saying so.

What this does to competitor research

You are sizing up a competitor. You copy an ASIN from the search results, which is whichever variation Amazon decided to show you that day, and your tool reports the keywords behind it.

If you landed on the colour that gets clicked, you see the product's real keyword footprint. If you landed on any of the others, you see a listing that appears to win nothing, in a niche that appears to be open.

The same mechanism distorts sales estimates, in the opposite direction. Amazon ranks the variation family as one thing but reports units sold per child. In our own data two products sit at rank 1 in the same subcategory, one with 8,000 units a month against its name and one with 50. They are not competing at different scales. One of them is a family where the clicks and the sales are recorded against a single child, and the other is a family where they are spread across a dozen.

Any estimate built on "rank predicts units" inherits that error, multiplied by however many variations the product has. It is the ordinary reason two well-known tools give you numbers three times apart for the same listing, and neither of them is lying.

Ranking, clicks and purchases are three different fights

Once the unit of measurement is settled, a second confusion is worth clearing. These three are not stages of the same thing.

A product can hold a strong organic position and take almost none of the clicks, because the two products above it take them all. It can take the clicks and lose the purchases: on the term mincer, one listing collects 7.3% of everything clicked and 1.2% of everything bought. Shoppers find it, choose it over the alternatives, open it, and then buy something else. That is not a visibility problem, and more advertising will make it worse.

Amazon publishes both shares separately, and it says plainly that they are independent measurements: purchase share can exceed click share, and often does on a product that converts well. Comparing the two is the useful part. A product taking a sixth of the clicks and a twentieth of the purchases is winning the search and losing the page.

Three names per phrase, and the rest of the market

There is one more consequence of "three per term" that changes how a keyword list should be read.

Because Amazon names exactly three, the shares on a phrase add up to how closed it is. Across the hundred terms behind one product we checked, the three most-clicked take between 9% and 54% of all clicks, with a median around 31%.

At the top of that range a phrase has owners, and a new listing does not join it by appearing in the results; it has to take clicks from a specific competitor. At the bottom, attention is spread thin enough that ranking is most of the work. The same position on those two phrases is worth completely different amounts, which is why a keyword list sorted by search volume alone routinely puts the wrong term first.

What to do with this on Monday

Look up the parent, not the variation you happened to copy. If the tool does not tell you which it used, assume the narrow answer.

When a competitor looks weak, check whether you are looking at their strongest child. A family whose clicks concentrate on one colour looks like nothing at all from any of the others.

Treat rank, clicks and purchases as three separate diagnoses with three separate fixes: position is advertising and relevance, click share is the image, the title and the price against the ones beside you, purchase share is the page itself.

And before spending on a phrase, find out how much of it the top three already hold. It is the difference between entering a market and attacking a position.

Figures throughout are from Amazon's Brand Analytics search terms data for the United States, July 2026.

You can run the same reading on any ASIN with our ASIN inspector: it merges the variation family, shows which terms Amazon names the product among the three most clicked for, and puts click share and purchase share side by side. The keyword tracker does the position half, and remembers where each phrase stood last time.