TL;DR
- Don't trust published accuracy numbers. They come from the vendors, use different definitions, and aren't audited.
- Test on a random sample of your own list (200 to 1,000 addresses), because lists differ by industry, source and age.
- The only real ground truth is a bounce. Send a small slice and count what bounces; compare it with each verifier's prediction.
- A fair test takes an afternoon and costs a few dollars.
Why vendor numbers don't help
When we searched for head-to-head accuracy data in 2026, almost every comparison was written by a verification vendor, and the figures conflict: the same product is quoted at 99%, 99.5% and 99.9% depending on who writes it. Accuracy also depends on what counts as correct. Is an unknown a miss? Is a catch-all deliverable? Each vendor chooses the definition that flatters it. A test on your own addresses removes that problem.
The test, step by step
- Sample. Take a random 500 addresses from the list you plan to send to. Shuffle first so the sample isn't just the oldest rows. Use our list sampler if you want a quick estimate before spending anything.
- Verify the same file with each tool. Keep the files identical, same order, same day. Free trials and small packs are enough; most tools give a few hundred free checks.
- Tabulate. For each verifier count deliverable, risky (catch-all, role, disposable), undeliverable and unknown. The unknown count tells you how often it gives up.
- Find the disagreements. Addresses where the tools disagree are the interesting ones. Put them in their own file.
- Send a small, real batch. From a warmed sending domain, send to 100 addresses that all tools called deliverable, and to every address one tool called undeliverable but another called deliverable. Don't send to addresses both called undeliverable.
- Count bounces after 48 hours. A hard bounce on a deliverable address is a false positive. A delivery to an address a tool called undeliverable is a false negative (a good lead lost).
What to compare
| Measure | How to calculate it | Why it matters |
|---|---|---|
| False-positive rate | Hard bounces / sends among addresses called deliverable | Drives your bounce rate and domain reputation |
| False-negative rate | Delivered / sends among addresses called undeliverable | Good leads you'd throw away |
| Unknown rate | Unknown / total | How often the tool gives up; unknowns are refunded by some vendors |
| Catch-all handling | Share flagged risky, and bounce rate if you send | The hardest category for every verifier |
| Cost per usable lead | Price / (deliverable count) | The number that decides the purchase |
Sample sizes matter: with 100 sends, one bounce is 1%. Treat anything under a few hundred sends as a rough signal, not a proof.
Pitfalls
- Testing on a list that is already clean, which makes every verifier look the same.
- Sending from a brand-new domain. A burned reputation hides real bounce differences. See how many bounces burn a domain.
- Counting soft bounces as failures. A full mailbox isn't a bad address; see hard vs soft bounces.
- Judging a tool on the first 50 results. Small samples swing wildly.
Run it on us
Mail Exchange gives you a free trial of 25 checks, which is enough to compare a handful of tricky addresses before you commit. For a bigger test, ask us. Whatever you pick, publish your own numbers to your team: it's the only accuracy data that applies to your list.
FAQ
How accurate are email verifiers?
Vendors state 98-99%+, but those are self-reported and defined differently. On your own list, real accuracy depends on how many catch-all and role addresses it contains.
How many addresses do I need to test?
200 to 1,000 random addresses gives a useful signal. Fewer than 100 can mislead.
What is a false positive in email verification?
An address a tool called deliverable that then bounces. It's the costly error because it hurts sender reputation.
Related
- How to choose an email verifier
- Verification statuses explained
- List sampler
- Cold email bounce rate benchmarks
Instantly
Smartlead
HubSpot
Mailchimp
Apollo
Zapier
Google Sheets
Make