Your catalogue calls it a category. Your buyer calls it a problem. The usual first move is to ask an assistant how you rank for the category name, and that question quietly assumes the buyer typed the category name, which is the part nobody checks. Frequently they did not. They described the job, in ordinary words, with a constraint attached, and the answer they got was assembled around that description rather than around your label.

This matters more in industrial and technical markets than anywhere else, because that is where the gap between the catalogue vocabulary and the buying vocabulary is widest. The catalogue was written by people who classify products. The question was typed by somebody who needs a thing that does something.

What follows is a procedure for finding out which words are actually in play. It takes an afternoon and needs nothing but a browser.

1. Write down the label, then write down the job

On one line, the name your catalogue uses. The internal one, the one on the navigation, the one your sales deck puts on the slide.

On the next line, the same product described only by what it does and what it has to satisfy, with no category noun at all. Not "industrial dosing pump" but "something that meters a corrosive fluid at a controlled rate in a food production line". Include the constraint that actually decides the purchase: the material, the tolerance, the certification, the temperature range, the throughput. That constraint is the part a specifier will not leave out.

Keep both lines. Everything below compares them.

2. Ask the job, not the label

Take the second line, phrase it as a question a person would ask, and put it to an assistant in a fresh session. Ask who supplies it, or what to use for it.

Then read the answer for two things. First, whether your category label appears anywhere in it. Sometimes it does not, and the answer is written entirely in different terminology, which tells you your label is an internal artefact rather than the market's word. Second, whether you are named. Those two findings are independent, and both are useful.

Run it 5 times, in fresh sessions. These answers vary between runs, and a single one will mislead you in whichever direction it happens to land.

3. Ask what the thing is called, in reverse

Now go the other way and ask the assistant directly what a product like the one you described is normally called, and what other names exist for it.

This is the step people skip and it is the one that pays. You will usually get several names: a formal one from a standard, a trade one, an abbreviation, and often one that is really a brand name that became generic in the way that happens to successful products. That list is an approximation of how the web talks about your category, which is the same material the assistant draws on when it answers a buying question.

Compare it against your own site. If three of those names never appear anywhere on your pages, a question phrased with any of them has less to match against you than against a competitor who uses them.

4. Do it again in each language you sell in

Category names are not translations of each other. A product family can have a conventional name in German that is not the German words for the English name, because the German-language trade settled on a different term, often an older one or one that came from a dominant supplier. Run steps 2 and 3 in each language, and expect the vocabulary lists to diverge rather than to line up.

This is also where a lot of exporters find their real problem. The local site was translated from the English one, so it carries the English category concept in local words, and the term the local market actually uses is nowhere on it.

5. Check the constraint, not just the noun

Go back to the constraint you wrote in step 1, and search your own site for it. The specific alloy, the standard designation, the tolerance figure, the approval.

Very often it exists, and it exists only inside a PDF specification sheet, as part of a table, sometimes as an image of a table. That is a page a fetcher can reach and cannot use. If the constraint that decides the purchase is not stated in text on a page, then a question built around that constraint has nothing on your site to work with, whatever your category page says.

6. Build the prompt set from what you found

Now write the questions you will measure with, and write them in the vocabulary the exercise produced rather than the one you started with. A working set has some of each: functional questions with a constraint and no category noun, questions using each of the alternative names you discovered, and a few using your own label, so you can see whether it does anything.

Then stop editing it. The comparison over time is the whole point, and a set you keep rewording produces movement that is entirely your own doing.

Automated prompt generation helps with the first draft and is one of the things PSentry does from your site, across 4 assistants and each language you sell in. It is worth being clear about what that can and cannot do: it works from what your site says, so it inherits your vocabulary. The alternative names in step 3, and the ones your local market uses instead of yours, are exactly the ones it has no way to know about. That part is yours, and it is an afternoon.