Wrong industry, wrong leads: how to check an audience before you pay for it
Most lead tools turn your description of your customers into industry codes with a keyword match, then bill you for whatever comes back. Here is how that fails, and the four checks that catch it before your money is gone.
Here is a failure this industry rarely writes about. A supplier of office supplies and toner, selling to legal and education firms, described part of his customer base as "care homes and clinics". A keyword matcher read the letters car at the front of "care", decided he meant car dealerships, and searched on that. Nine of his 25 leads were car dealers. He had paid for all of them.
That is not an exotic bug. It is the ordinary behaviour of every tool that turns your words into industry codes by looking them up rather than understanding them. This post is about how to spot it before your budget goes, whoever you buy from.
Why keyword matching fails on real sentences
A lexical matcher has no idea what your business is. It has a list of industry names and a string from you, and it looks for overlap. That produces three predictable classes of mistake:
- Prefix collisions. Care and cars. Hospital and hospitality. Light and lighting. The letters agree and the meanings do not.
- Context blindness. "Pharmaceutical" from an office supplies seller means pharmaceutical companies who buy paper. A matcher can just as easily land on the growing of pharmaceutical crops, because it never asked what you sell.
- Dumping grounds. Vague phrases like "general business" get pulled toward whichever category happens to share a word, and the public record has plenty of catch-all categories waiting to absorb them.
None of these look wrong on a summary screen that says "12 industries matched". They look wrong when the leads arrive, which is after the money has moved. A bought list has the same problem one step further back, as our comparison of buying, building or generating a B2B list sets out: you inherit somebody else's classification and never see how it was made.
The AI reads your whole business rather than one keyword, shows you exactly what it will search, and lets you change any of it in one tap. 25 free leads, no card.
The four checks worth demanding
Whatever you are buying, these four questions separate a vendor who can show its working from one who cannot.
1. Show me the categories before you charge me
You should be able to see the exact industry categories that will be searched, the geography, and ideally real sample companies from that search, before a single credit is spent. "Trust us and check the output" is a business model, not a safeguard.
2. Tell me when you are not sure
The dangerous answer is not "no match", it is a confident wrong match. A tool that says it could not match with confidence and leaves the phrase out has told you something useful. A tool that silently picks its best guess has handed you a bill and a problem.
3. Let me fix it in one tap
Seeing a wrong category is only half a safeguard. You need to remove it, swap it or add the right one immediately, without an email to support and without rebuilding the search.
4. Check the results against what I said, before delivery
The audience can be right and the rows still drift, because industry codes are self-declared and companies file more than one. The check that matters is the one that happens after the search and before the leads land.
What we changed, and why
We shipped all four on 18 August, on the back of exactly the case above.
The AI reads your whole business, not one keyword. Your trade, what you sell, how you describe your customers, your best customer, who is a bad fit. It picks register categories in that context, and it can only pick from categories genuinely related to the phrase you wrote. Care homes can never become the sale of new cars, however the letters line up.
Low confidence is shown as low confidence. Anything the model is not sure of is either offered as a suggestion you tick yourself, or reported honestly as not matched. Nothing uncertain is searched on your behalf.
You confirm before you spend. The accept screen shows the categories, the area and real sample companies, and every part of it is changeable in one tap.
Results are re-checked before delivery. After the run and before anything reaches your dataset, the rows are compared against your own description of your business. Rows that do not match are held back. A held-back row is never delivered and never charged, and the unspent credits stay yours.
Every lead row shows its category. So you can audit the search yourself in a glance rather than taking our word for it.
How to run this test on any vendor
You do not need our product to apply the standard. Take one phrase from your own business that is genuinely ambiguous, the sort of thing a human would ask a follow-up question about, and give it to the tool you are considering.
- Write the phrase as you would say it. "Care homes and clinics", "practices", "independent retailers", "trades".
- Ask to see what it will search. If there is no screen between your sentence and your money, that is your answer.
- Look for the ambiguous ones. If your phrase resolved into a suspiciously specific category, ask why.
- Check the first ten rows against your own definition of a customer. Not the first hundred. Ten is enough to see a systematic miss.
- Ask what happens to a wrong row. Refunded, replaced, or your problem?
Codes are a starting point, not the answer
Even a perfect match to an industry code leaves work to do, because companies self-declare their codes at incorporation and often never revisit them. That is a whole subject of its own, and which SIC codes actually describe your customers covers how to choose them and what they cannot tell you. The complete guide to SIC codes has the structure behind them. Our methodology page sets out where every field on a lead comes from, and the SIC code hub has a page per code with the live company count behind it.
The honest version of the promise
We are not going to claim this can never go wrong. An industry classification built from self-declared codes will always have edges. What changed is where the cost of an edge case lands: on us, in a held-back row you were never charged for, rather than on you, in nine car dealerships you paid full price for.
Start free with 25 leads and no card, put your own awkward phrase into it, and read the confirm screen before you spend anything. That screen is the product.
Common questions
Why do lead generation tools return the wrong industry?
Most convert your description into industry codes by matching words rather than understanding the sentence. Prefix collisions like care and cars, and context blindness like pharmaceutical crops for an office supplies seller, both produce confident wrong matches that still get charged.
Can I see which industries will be searched before I am charged?
In Leadistry, yes. Before any credit is spent you see the exact register categories, the area and real sample companies from that search, and you can change any of it in one tap.
What happens to leads that do not match my business?
After the search and before delivery, the rows are checked against your own description of what you sell and who buys it. Rows that do not match are held back. They are never delivered and never charged.
How can I test a lead vendor for this problem?
Give it one genuinely ambiguous phrase from your own business, then ask to see what it will search before you pay. Check the first ten rows against your own definition of a customer, and ask what happens to a row that misses.