We publish 74% as the share of a normal B2B search that comes back with an address, and it is a real number from real runs. It is also going to be wrong for you, possibly by twenty points in either direction, and the reasons are worth understanding before you use anybody’s headline figure — ours included — to plan a quarter.
A hit rate is a fraction. Almost all of the argument is about the bottom half.
One run, five honest numbers
Take a run of a thousand people from a search page. Suppose it goes like this, and the numbers are made up so that the arithmetic is checkable rather than impressive:
- 1,000 rows on the search.
- 60 of them have no employer listed at all — self-employed, stealth, a profile that names no company. Nothing can be attempted for these.
- 940 attempted.
- 740 people come back with a personal work address.
- A further 55 come back with only a role mailbox:
info@,sales@,contact@. Nobody’s name is on it.
Now report it.
- 740 / 1,000 = 74%. Per person, personal addresses only, every row in the denominator. The strictest version, and the one we quote.
- 795 / 1,000 = 79.5%. Role mailboxes counted as hits.
- 740 / 940 = 78.7%. Rows that could not be attempted removed from the denominator, on the argument that they were never a fair test.
- 795 / 940 = 84.6%. Both of the above.
- 740 / 1,850 = 40%. Per lookup rather than per person — the engine tried roughly two candidate addresses per person, and each attempt counts.
Five numbers, one run, no dishonesty anywhere. The spread is 45 points, larger than the gap between any two vendors in this category. Comparing published hit rates is close to meaningless unless each one states its denominator, and most do not.
If you are evaluating something, the question to ask is not “what is your hit rate”. It is: per person or per lookup, do role mailboxes count, and are the rows you could not attempt in the denominator?
What counts as a hit is also a choice
Beneath the denominator argument there is a quieter one about the numerator.
A role mailbox is not a person. info@ reaches a shared inbox, sometimes a
ticketing system, occasionally nobody. As an outbound target it is worth close
to zero, which is why we do not count it as a hit and do not charge for it. A
tool that counts it has a different definition, not a worse product — and a hit
rate three or four points higher.
A single-source address is a claim, not a confirmation. If an address has been seen once and never corroborated, counting it as a hit assumes you will send to it. Some teams will and some will not, and on catch-all domains the distinction matters most — see what a catch-all does to your send. A confirmed-only hit rate is always lower than an anything-returned one, and both are defensible.
An address that exists but the person left. Common, and rarely counted against anyone: the mailbox may still accept mail while the human is three months into another job.
The right address for the wrong person. Two people with the same name at similar companies. Rare, embarrassing, and it counts as a hit in every measurement scheme there is.
What actually moves the number
Composition does more work here than tooling. Five factors, roughly in order of how much they matter:
Whether the employer has a public domain. The single biggest one. A person at a company with its own website and its own mail domain is findable. A freelancer on a personal Gmail, someone at a company that runs everything through a marketplace, a profile that lists no employer — these are not hard, they are impossible, and a search full of them will halve any hit rate. It is also why a search filtered by headcount behaves so differently from one filtered by title alone.
Company size. Very small companies publish little and often have one mailbox for everything. Very large ones publish plenty but sit behind filtering gateways that accept every address, which makes confirmation harder even where coverage is good.
Industry. Software, agencies, consultancies and professional services put addresses in public: conference lists, press pages, code repositories, filings. Construction, hospitality, retail and the trades often publish a contact form and nothing else. Healthcare and education have addresses in abundance, behind directories that are not open.
Region. Publication norms differ by market. Where fewer personal addresses are published, where shared inboxes are the convention, or where the legal culture discourages the sort of publication this depends on, any tool’s rate falls. That is a property of the market rather than a tooling gap, and it is the reason a rate quoted without a region attached tells you very little.
Seniority and tenure. Founders, partners and VPs are published because publishing is part of the job. Someone six weeks into their first role is not published anywhere yet, and someone who changed jobs last month has a live mailbox nobody has written down. Recency cuts both ways: it is why a search of recent joiners scores low and why job-change monitoring is worth having.
Put those together and the same engine, unchanged, does far better on a search of software VPs at mid-market US companies than on one of independent restaurant owners. Neither result tells you anything about the tool on its own; both tell you something about the list.
Why a vendor’s number can never be your number
A published hit rate is an average over everything everyone ran, weighted by whoever ran the most volume: a statement about a population you are not in. Ours is measured per person, personal addresses only, misses in the denominator — the strictest of the five framings above, and still an average of other people’s lists. The only figure that can inform a decision is measured on your own ICP, and that is a short piece of work.
Measuring your own, in one afternoon
- Pick a real search, not a flattering one. The one you would run on a Monday. If your outbound covers three segments, run each one separately — a blended number hides the one that is dragging.
- Use enough people. Two hundred rows puts the answer within about six points either way at 95% confidence, which is enough to choose a tool or plan a quarter. Fifty rows can be twelve points out in either direction and will tell you a story you then act on for a year. The free plan is 50 addresses a month, which is enough to check the shape of the output and not enough to measure a rate.
- Write the denominator down before you run. Decide now whether rows with no employer are in it and whether role mailboxes count. Deciding afterwards is how a measurement becomes a justification.
- Run it and record four counts. People attempted, people with a personal address, of those how many were corroborated by more than one source, and people with only a role mailbox.
- Send, then wait two weeks and count twice more. Hard bounces, and complete silence at the account level — a company where nobody replied at all is worth a look, because that is what a catch-all does to a segment.
- Compute two rates. Find rate is people with an address over people attempted; reach rate is people who received the mail over people attempted. The second is what your pipeline is built on, and always the lower.
- Re-run the same search in a month. The difference is your list’s decay rate, and it is the number that decides how often the whole exercise needs repeating.
Two practical notes. A miss costs nothing and a role mailbox costs nothing, so a measurement run charges only for the people it finds — which is also the arithmetic in the price per email you can actually send. And if the number comes back low, look at the composition of the list before the tool: nine times out of ten the answer is in the filters.
Hit rate is not really a product feature. It is a property of the list you point the product at, and the only honest way to publish one is to say what was divided by what.