B2B data

Count the market before you pay to collect it

Deciding which companies to approach is a question about your market, and the software only writes the answer down. Two shapes exist. You describe the market with filters, industry, headcount, country, technology, and you get back what your theory predicted. Or you hand over the accounts that already bought and ask what resembles them, which needs a seed of real customers and is worth nothing without one. Where the second one stops working is published in targeting by resemblance instead of by criteria.

The expensive mistake at this step is not the subscription. On one of our runs, four searches returned 23,089 rows that resolved to 15,363 distinct people. A third of that spend bought nothing, because the searches were ordered by how they came to mind instead of from narrowest to widest, and a duplicate is billed the moment it comes back. Deduplicating afterwards refunds none of it.

Counting is usually free and collecting never is, which is the whole method in count before you buy the list. On Airscale a search bills 0.1 credit per row returned whether you keep the row or not, and the total field of that paginated response comes back as zero, so the free count is the only number you can plan against.

Before signing anything, test the coverage on twenty companies you know intimately and count what comes back right. A database that is excellent in the United States and thin in Europe is a good product and a bad choice, if you sell in Europe. The tools we pay for at this step are on B2B data and targeting.

Order the providers cheapest first, and verify before you send

A name and a company go in, an email address and a mobile number come out. What that costs depends on three things, and the headline rate is none of them.

Billing shape decides the most. Prospeo charges ten credits for a mobile and one for the email that arrives with it, billed only when something is found, which their own pricing FAQ states in writing. A provider that bills per row queried charges you for every miss, and on European mobile data the misses are the majority of the file.

Then the hit rate on your own market, which no vendor page can answer for you. Our waterfall on 8,948 French decision-makers in August 2026 returned 1,798 mobiles. Twenty percent, and that is the whole waterfall rather than any single provider. Anyone promising materially more than that on European decision-makers is selling you a number and not a contact.

Then the order you call them in. Cheapest first, stop at the first hit, and let the expensive one meet only the rows nobody else could find. Backwards, the same data costs around forty percent more and the invoice never tells you why.

Verification sits under all of it and gets skipped because it looks like a rounding error. On Icypeas it is a tenth of a credit, ten times less than the number people assume, and that gap is the difference between verifying the whole file and verifying only what looked doubtful. Two years of running one in production are in what a mobile really costs, and the side by side is on Email and phone enrichment.

Slice a large search yourself, or buy the same people twice

Extraction is the dullest step of the three and it sets the ceiling on everything after it. A search has to become a clean file. Three things go wrong, and it is always the same three.

Volume. Above a couple of thousand profiles a search will not come out in one piece, so someone slices it by hand, and hand-slicing is how overlap gets created without anyone seeing it.

Overlap, again, because it is the same money as the step above. Two searches that feel different to a human return the same people to a database. Ours cost a third of a run, and the ordering that prevents it is in scraping a market without paying twice.

Field hygiene, which sounds cosmetic right up to the moment it is not. An emoji inside a first name, a job title glued to a surname: both survive the extraction, travel through the enrichment untouched, and land in the opening line of an email in front of a prospect. Vayne strips them on the way out, and its free plan gives 200 leads a month, which is enough to see whether the output matches what you actually need before paying for it.

Judge a tool here on what it does without being asked. Split a large query on its own. Recognise profiles you already pulled. Clean the fields. Tell you the count before the run starts. Prices go by volume of profiles per month and the ladder usually opens around thirty dollars. What we use is on LinkedIn scraping and extraction.