How to Evaluate a B2B Data Provider in 500 Records
Every data vendor will show you a demo search that works. The only evaluation that means anything uses records where you already know the answer.
Every contact data vendor can run a demo search that returns impressive results, because they choose the search. The only evaluation worth anything uses records where you already know the correct answer, on the segment you actually sell into.
Here is that test, and the questions that go around it.
The 500-Record Test
Assemble 500 contacts where you have a verified email, existing customers, closed-won deals, anyone you have successfully corresponded with. These are your ground truth.
Strip the emails. Send the vendor name, company and title only. Then measure four things against what comes back:
- Match rate, how many records they returned anything for.
- Accuracy, of those, how many match the address you already had.
- Freshness, how many of their answers reflect a job change you know about.
- Confidence calibration, when they flag an answer as high confidence, is it actually more often right? A confidence score that does not correlate with accuracy is decoration.
That last one is the most revealing and the least asked about. A provider returning 80% coverage with meaningless confidence scores is harder to use than one returning 60% with honest ones, because you cannot decide what to trust.
Test the Hard Half
One refinement that changes the result. Split your 500 into the easy half (large, well-documented companies) and the hard half: small, regional, recently founded, or in an industry nobody writes about.
Everybody performs on the easy half. The hard half is where providers differ, and it is usually where your actual growth segment sits.
If a vendor is strong on the easy half and weak on the hard half, that is not necessarily disqualifying. It means they are a good first pass in a waterfall rather than a single solution, which is useful to know before you price them as the latter.
Six Questions for the Call
- Where does the data come from? Purchased, licensed, contributory, collected? Vendors that will not answer specifically are usually aggregating someone else's work.
- How often is a given record re-verified, and can you show me the last-verified date on a record? "Continuously" is not an answer.
- What is your coverage on my ICP specifically, not your database total? Ask for the number on the 500-record test.
- What happens to unused credits at renewal? Roll over, expire, or partially roll?
- Does exported data remain mine after termination? Get this in writing.
- What is the renewal uplift cap? If there is none, assume the worst, because this is where multi-year regret originates.
Compliance Is Not Optional Diligence
Ask how the data was collected and what lawful basis applies to the contacts you will receive, particularly for anyone in the EU or UK where GDPR applies to business contact data too.
Ask what their process is when a person asks to be removed, and how quickly that propagates to you. Suppression that takes a quarter to reach your sending tool is a problem you inherit.
This matters practically as well as legally: a list that generates removal requests generates complaints, and Google's threshold is roughly one complaint per thousand delivered before your sending reputation suffers.
Price It Per Usable Contact
Headline pricing is per credit or per seat. Neither is the number you care about.
Divide the annual cost by the number of contacts you will actually be able to send to, match rate times accuracy times your verification pass rate. A provider at half the price with a third of the usable coverage is more expensive.
Then compare that figure against what a contact is worth to you: average deal value times the rate at which a sequenced contact becomes a deal. If cost per usable contact exceeds that, no amount of coverage makes it a good purchase.
Before You Sign Anything Annual
Run the test as a paid pilot. Vendors will resist and most will agree, and a vendor who refuses a 500-record test on your own data has told you what the test would show.
Start with the shortest term available even at a worse rate. The discount for a two-year commitment is almost never worth being locked to a provider whose coverage on your segment you have measured exactly once.
Frequently asked questions
How do I test a B2B data provider before buying?
Take 500 contacts where you already have verified emails, strip the addresses, and send only name, company and title. Measure match rate, accuracy against what you knew, freshness on known job changes, and whether their confidence scores actually correlate with accuracy.
What should I ask a contact data vendor on a sales call?
Where the data comes from, how often records are re-verified and whether they can show a last-verified date, coverage on your ICP specifically, what happens to unused credits, whether exported data stays yours after termination, and the renewal uplift cap.
How should I compare data provider pricing?
Per usable contact, not per credit or seat. Divide annual cost by match rate times accuracy times your verification pass rate, then compare that to what a contact is worth, deal value times the rate at which a sequenced contact converts.
Should I sign a multi-year data contract for the discount?
Rarely. The discount is almost never worth being locked to a provider whose coverage on your segment you have measured once. Start with the shortest term available, even at a worse rate.
Want help putting this to work?
Talk to a grobot strategist about wiring this into your stack.
Talk to a Strategist →