Home › Blog › MillionVerifier vs ZeroBounce vs NeverBounce: Run Your Own Accuracy Test
MillionVerifier vs ZeroBounce vs NeverBounce (and BounceBan vs LeadMagic on Catch-Alls): How to Run Your Own Accuracy Test
Verifier accuracy claims are unfalsifiable marketing until you test them on your own data, and the vendors disagree with each other constantly, especially on catch-alls, which is exactly where bounces come from. Here is the frozen-list test protocol that measures real accuracy in two weeks, and what the disagreements teach you about building your waterfall.
Quick Answer: Freeze a List, Let Bounces Grade the Vendors
Published accuracy numbers are marketing; your bounces are data. Freeze 1,000+ real ICP rows, run the identical list through each vendor in one week, then send and record actual hard bounces. Score false valids (bounced despite "valid", the damaging miss), false invalids (discarded pipeline), and unknown rates. For the catch-all specialists, compare commitment rate and cost per recovered address on the ambiguous slice only. Outcomes pick the waterfall seats; re-test quarterly.
The Protocol
WEEK 1 freeze list: 1,000+ rows, your real ICP, cleaned first (CSV Cleaner) so vendors are graded on verification, not hygiene → same list to each candidate, same week → record verdict per row per vendor WEEK 2 real campaign to the list (normal copy, isolated segment, stricter stop rule) → log hard bounces per address SCORE per vendor false-valid rate said valid → bounced false-invalid rate said invalid → provably fine unknown rate refused to commit cost per settled row CATCH-ALL ROUND (BounceBan / LeadMagic class) only the catch-all rows → commitment %, accuracy vs bounces, cost per recovered address
- •False valids are the error that matters most: they convert directly into hard bounces on your domains. A vendor with a slightly higher price and a meaningfully lower false-valid rate wins the stage-1 seat every time.
- •Unknown rate is a real cost, not a neutral: every unknown either gets discarded (pipeline loss) or escalated (specialist spend). Two vendors with equal accuracy and different unknown rates are not equal.
- •Test with your list's provider mix, because verifier performance varies most against Microsoft-hosted and gateway domains; a test list that under-represents your enterprise share flatters everyone.
- •Database badges are not verification: Apollo-exported "verified" rows belong in the test as ordinary unverified input, the badge post shows why.
Reading the Results Into a Stack
The typical outcome: two bulk vendors within a point of each other on false valids, separated mainly by unknown rate and price, and the specialists meaningfully apart on commitment rate for catch-alls. That maps directly onto waterfall seats: cheapest acceptable false-valid rate takes stage 1, best cost-per-recovered-address takes stage 2, and the loser of each round becomes your quarterly challenger. Resist crowning a forever winner, vendor infrastructure changes, server blocking changes, and the quarterly mini-test is what keeps the stack honest without re-running the full protocol.
Frequently Asked Questions
Which email verifier is most accurate?
Nobody can answer that for your list, and that is the honest answer: accuracy varies by the mix of providers and catch-all share in YOUR data, vendors' infrastructure gets blocked differently by different mail servers, and published accuracy claims are self-reported. What is consistently true: the big bulk services (MillionVerifier, ZeroBounce, NeverBounce) agree on the easy 70-85% of rows and diverge on catch-alls and unknowns, and the divergence is where your bounces live. So run the frozen-list test on your own data and let outcomes pick your vendors.
How does the frozen-list accuracy test work?
Freeze one representative list (1,000+ rows from your real ICP, cleaned first), run the identical list through each candidate service in the same week, and record every verdict per row: valid, invalid, catch-all, unknown. Then the ground truth pass: send a real campaign to the rows and record actual hard bounces per address. Score each vendor on false valids (said valid, bounced, the damaging error), false invalids (said invalid, another vendor's valid verdict plus delivery proved otherwise, the lost-pipeline error), and unknown rate (how much they refused to commit). Two weeks, a few hundred dollars, permanent clarity.
What do verifier disagreements actually look like?
On typical B2B lists, expect vendors to agree on roughly 80% of rows and split three ways on the rest: one calls it valid, another catch-all, a third unknown. The disagreement concentrates in catch-all domains, greylisting-heavy servers, and gateway-protected enterprises, exactly the segments where guessing wrong costs bounce rate or discarded pipeline. This is the empirical case for the waterfall: no single opinion is trustworthy on the hard slice, so route the hard slice to a specialist.
BounceBan or LeadMagic for catch-all verification?
Both exist specifically for the ambiguous slice, and the same protocol settles the choice: send both services ONLY your catch-all rows from the frozen test, then compare (1) commitment rate, how many rows each is willing to call deliverable or not, (2) accuracy of those calls against the real-send bounce outcomes, and (3) cost per recovered address, the deliverable-and-did-not-bounce rows you would otherwise have discarded. Vendors shift over time as their infrastructure evolves, so re-run the mini-test quarterly rather than marrying a result.
How often should I re-test my verification stack?
Quarterly mini-tests (a few hundred rows through the incumbents plus one challenger) and a standing scorecard: every campaign, log bounce rate by verification verdict. If rows your stage-1 vendor called valid start bouncing above ~1%, or your specialist's recovered catch-alls exceed the 3% segment stop, the stack is drifting and the test tells you where. The scorecard costs nothing, it is the same campaign data you already collect, read against verdicts.
Get FREE .com Domains
Prewarmed Inboxes So You Can Send Cold Emails Today
Get Prewarmed & Fresh InboxesPrewarmed & Fresh Inboxes
Google, Microsoft 365 and Azure inboxes with free prewarmed domains and DNS configured before you log in. Prewarmed inboxes send the day they land; every order is delivered within 48 to 72 hours.
Get Prewarmed & Fresh Inboxes