One pricing page in front of you offers 10,000 credits a month. The tab next to it offers 2,000. Before you start dividing, notice that nobody has defined the unit. There is no weights-and-measures office for sales data. Every vendor mints its own credit, decides what that credit buys, and can revise the definition in a pricing page edit on a Tuesday afternoon. Those two numbers are denominated in two different currencies and the exchange rate is not published anywhere.
The only question worth your attention is what a single credit resolves to when it touches the ground.
A credit should buy you a fact you did not already have
Strip the packaging off and a data vendor has three things it can genuinely sell: a name attached to a company and a role, a phone number that rings that person, and an email address a mail server will accept. Those cost real money to collect and verify, and they rot on the shelf whether anyone buys them or not. Metering them is fair.
Searching is not one of those three. Filtering by headcount and geography is not one of them. Counting how many companies match your criteria is not one of them either. That work costs the vendor a database query and costs you an afternoon of judgment, which is your job and which you are already paying for with your own time. Charge for the query and what you are really charging for is the right to find out whether the inventory is worth anything.
This is the first place a second meter hides, and it is the oldest trick in the category. Reveal-gated interfaces are everywhere: you get the shape of a person, blurred, and you spend to learn who it is. Lusha sells packs of credits you burn to reveal contact details. The design points one direction, which is that you commit before you know.
We built ours the other way. Searching is free. Counting the market is free, and I think that is the most underrated thing on our free list, because finding out a segment holds 40 people rather than 40,000 changes what you do with your entire quarter. Previewing 25 of those people by name is free too. Read two dozen real names, look at the titles, decide the list is garbage, walk away having spent nothing.
Paying for the attempt is a different deal from paying for the answer
Waterfall enrichment sends a record down a line of data providers until one of them returns something. Clay sells that arrangement and plenty of others have copied the shape of it. The billing question underneath is the one buyers skip past: when the record walks the whole line and comes back empty, who pays for the walk?
Under per-attempt pricing, you do. You are buying effort rather than information. And the accounts where enrichment fails are disproportionately the small, strange, under-documented companies nobody has indexed yet, which in a lot of markets is precisely the segment worth selling into. You spend the most on the prospects you learn the least about.
Our version: adding a person to your pipeline costs one credit. A verified phone and email cost five, and we charge those five only when we actually find them. A miss costs you nothing. I would not call that generosity. It is what the word credit is supposed to mean.
Three balances against one person, and the smallest one is your real plan
Watch for vendors running more than one meter at a time. The common setup is a pool of email credits, a separate and much smaller pool for mobile numbers, and then an export allowance sitting on top governing how many records you can take out of the system at all. Apollo publishes credit allocations and export limits that vary by plan. Multi-pool pricing is the category norm, not some outlier.
What that does in practice is set your capacity by the scarcest balance, which is never the generous one printed in the largest type. You can be sitting on thousands of unused email credits and still be unable to make a single call, because the mobile pool emptied in week two.
The export line deserves its own suspicion. A fee for removing your own list is a switching cost wearing a fee's clothing. It exists because your leaving is bad for the vendor.
The expensive double-metering starts after you leave the data tool
Say you get a clean contact at a fair price. You have not finished paying for that human being.
The sending platform meters them again, usually by contact or by active lead. The CRM counts them a third time as a billable marketing contact. The dialer bills a seat and then bills the minutes. And if you run an answering service so the callback does not drop into voicemail, that meter runs too. Ruby and Smith.ai both price answering by the call or by the minute, which means a heavy week of inbound is also an expensive week, and the weeks you most want the phone ringing are the weeks the invoice climbs.
Same prospect, four vendors, four meters, each one attached to a different verb applied to the same person. The bill roughly doubles somewhere between finding someone and speaking to them, and the founder never sees that doubling in one place, because it shows up on four separate invoices in four different weeks.
Our meter stops at the data. Writing to those people, reading what comes back, sorting the replies, answering your number at eleven at night when someone finally calls: none of it is priced per contact. Your tier sets how many AI workers you have, not how many humans you are allowed to talk to. Launch is $24.99 with 100 credits a month, Frontier $79 with 250, Mission Control $149 with 500.
The arithmetic out loud, including the unflattering part
A fully enriched contact, added with a verified phone and email found, runs six credits. On Launch that works out to sixteen of them in a month. I would rather write that sentence than bury it under a big round number.
Sixteen is still the wrong way to read the allowance, because enriching everybody is the wrong move anyway. Add 100 people for 100 credits. Read them. Most of them you are never going to call. Spend the extra five only on the ones you would genuinely pick up the phone for. The meter stays under your thumb at every individual decision rather than running quietly in the background against a list you bought blind.
Four questions to put to any vendor before the card goes in
- Can I see and count the segment before spending anything, by name, unblurred?
- When enrichment fails, do I pay for the attempt?
- How many separate balances are there, and which one runs out first?
- What else gets metered after the data: the sending, the answering, the export?
Metering data is defensible. It costs money to acquire and it decays on its own schedule. What is not defensible is metering the parts that cost the vendor nothing, then charging again for the same person at every verb until you are paying four separate rents on one prospect who has not replied yet.