Every contact data vendor quotes a price per credit, and every buyer repeats it back as if it were the price of an email address. It is not. It is the price of an attempt at one. The two numbers are separated by your hit rate, your role mailboxes, your duplicates and the rows you would never send to, and by the time you have divided the bill by the people you could actually contact, the figure has moved enough to change which plan you should be on.
Here is the arithmetic, on a list of a thousand.
Four denominators, and only one of them matters
Take the same monthly invoice and divide it four ways.
- By credits bought. The vendor’s number. It is the largest denominator, so it produces the smallest price, which is why it is the one on the page.
- By people looked up. Lower, if anything you looked up came back empty and was still charged for.
- By addresses returned. Lower again. This is the first honest figure.
- By addresses you would actually put in a sequencer. Lower still, after role mailboxes, duplicates against your CRM, people who have left, and the single-source rows you decided against.
The gap between the first and the last is the whole subject. Nobody is lying to you at step one; they are simply answering a question about their inventory when you were asking one about your pipeline.
A thousand people, priced twice
Say you run an ICP search that puts a thousand people in front of you, and around 74% of them come back with an address, which is what a normal B2B search does. That is 740 people you can write to and 260 you cannot.
Both columns below use exactly the same credit price — Cuvy’s Pro plan, so a credit costs $0.0098 either way. The only thing that changes between them is which events count as chargeable.
| Row | Billed per person | Billed per attempt |
|---|---|---|
| People on the list | 1,000 | 1,000 |
| Chargeable events | 740 people found | 1,000 lookups |
| What that costs | $7.25 | $9.80 |
| Addresses you can send to | 740 | 740 |
| Price per sendable address | $0.0098 | $0.0132 |
| A person nobody could find | Free | Charged |
| A role mailbox instead of a person | Free | Usually charged |
| A second address for the same person | Free | Charged again |
| Running the same person again next month | Free | Charged again |
Same list, same addresses, same tool quality, and the price per address you can send to moves from $0.0098 to $0.0132. That is a third more, and it is invisible on any pricing page, because it is not a property of the price. It is a property of the denominator.
Two things follow that are worth saying plainly. The first is that per-attempt billing gets more expensive the harder your list is: at the hit rate above the penalty is about a third, and on a list where only half the people can be found it is double. The second is that the penalty is paid precisely when the tool performs worst, which is a strange incentive to hand a vendor.
The line items on nobody’s invoice
Bounces are paid in reputation, not in dollars
A bad address does not cost you the credit you spent on it. It costs you a fraction of your domain’s standing with the mailbox providers, and that account is not itemised. Send enough mail to people who do not exist and the effect shows up as reduced inbox placement for the people who do — including the prospects who were never going to bounce and the customers who already reply to you.
Nobody outside Google and Microsoft knows the exact threshold, and neither publishes one. The usual advice is to keep hard bounces to a couple of percent and treat that as a ceiling rather than a target. What matters for this arithmetic is that the cost is real, deferred, and charged to a different budget from the one that bought the addresses.
Cleaning takes as long as it takes, and it recurs
Work out your own version: the minutes it takes to open a file, sort it, drop the role mailboxes, spot the duplicates, look up the four rows that seem wrong, and re-export it. Call it forty minutes for a thousand rows. Two lists a month is sixteen hours over a year of doing this, and sixteen hours of somebody who does outbound is worth more than the software is.
That time does not disappear because a vendor is cheap. It disappears when the file arrives sorted — when the addresses two sources agree on come out in one file and the single-sighting ones come out in another with their source attached, so the judgement call is presented rather than discovered.
Credits you paid for and did not use
The other end of the same problem. A plan whose allowance you use half of doubles your real price per address, and the honest version of this argument cuts against us as much as anyone.
Pro is $0.0098 an address at its full 5,000. Use 2,000 of them in a month and you have paid $0.0245 an address, which is worse than Growth’s list rate of $0.019. That comparison is a trap, though, because Growth’s list rate only runs as far as its 1,500 allowance. Past that you are buying packs, and packs cost more than any plan. Growth plus the smallest pack covers 1,750 addresses for $44, still under Pro. One address more and you are into a second pack at $59, which is above Pro’s $49. So the real crossover is around 1,750 addresses in a month: below that, the smaller plan genuinely is cheaper, and above it Pro is, whatever the sticker rates say.
What softens it is rollover. Unused monthly credits carry into the next month up to twice the allowance, so a quiet month is deferred rather than lost — up to a point, and that point is 2×. Top-up packs never expire at all, though they are the most expensive way to buy: 250 for $15 works out at $0.06 an address, six times the Pro rate. Packs are for the month you run over, not for the year you run on.
Seats, if the tool charges for them
Per-seat pricing puts a second variable in the denominator: the price per address now depends on how many people are logged in, and a team of four with one heavy user pays four times for one user’s work. Worth checking which of the two numbers you are actually being sold.
Work out your own in ten minutes
- Take last month’s invoice. One number, the total.
- Count the distinct people you actually sent to from data that tool produced. Distinct people, not rows, not sends, not sequence steps.
- Divide. That is your real price per contactable person.
- Compare it with the per-credit price you thought you were paying. The gap between the two is your hit rate and the share of the file you discarded, converted into money.
Then do the same for the tool you are considering, using its billing rules and your hit rate rather than its. If you do not know your hit rate on your own ICP, that is an afternoon’s work and it changes every number above.
One last thing worth checking on any quote: what happens when the allowance runs out mid-run. An overage rate turns an unknown quantity of work into an unknown quantity of money, which is the one line item you cannot put in this arithmetic at all. Cuvy’s answer is that there is no overage — the run pauses and asks — and the full set of rules is on how we bill, beside the plans.