Key takeaways
- Accuracy claims are measured by vendors on their own test sets; they are not comparable across tools.
- The real question is: how many verified "valid" addresses bounce when you send?
- Catch-all, greylisting and protected servers create uncertain results for every tool.
- Tools differ mainly in how they classify uncertain addresses, not in obvious invalid ones.
- Test on a sample of your own list, and judge by bounces after sending.
In this guide
- What "accuracy" really means
- The easy cases and the hard cases
- Why two tools disagree
- The limits every verifier shares
- Deliverability guarantees
- How to test a verifier on your own data
- Reading your results
- What accuracy is worth
- Questions to ask a vendor about accuracy
- Accuracy and timing
- A simple way to track accuracy over time
- Glossary of accuracy terms
- What to do with the results of a comparison
- Summary
- Keep results accurate over time
- Frequently asked questions
What "accuracy" really means
When a verification tool advertises "99% accuracy", it rarely tells you what was measured. Accuracy can mean very different things:
- The share of addresses classified correctly on a test set chosen by the vendor.
- The share of addresses marked "valid" that do not bounce when you send.
- The share of all addresses that get a definitive answer (valid or invalid) rather than "unknown".
These are different numbers. A tool can be very accurate on what it classifies and still return many unknowns. Another can classify almost everything and make more mistakes on the hard cases. For a sender, the practical question is simple: when I send to the addresses marked valid, how many bounce?
The easy cases and the hard cases
| Case | Difficulty | Why |
|---|---|---|
| Bad syntax, non-existent domain, no MX | Easy | Pure technical checks, every tool agrees |
| Known disposable domains | Easy to medium | Depends on how up to date the tool's data is |
| Mailbox clearly exists or clearly rejected | Medium | Requires a good SMTP infrastructure and retries |
| Catch-all (accept-all) domains | Hard | The server accepts everything; see catch-all emails |
| Greylisting, timeouts, rate limits | Hard | Temporary refusals lead to "unknown" |
| Large providers that protect against probing | Hard | Tools rely on several signals and their own data |
Almost all differences between tools happen in the hard cases. On the easy ones, any serious verifier gives the same answer.
Why two tools disagree
- Infrastructure: the reputation of the servers used for checks affects how mail servers respond.
- Retry strategy: some tools retry greylisted addresses more patiently.
- Classification rules: one tool marks an uncertain address "risky", another "unknown", another "valid".
- Data: lists of disposable domains, known traps and historical results differ.
- Timing: verifying the same list a week apart gives slightly different results, because mailboxes change.
The limits every verifier shares
No tool can be 100% accurate, for reasons outside its control:
- Mailboxes change after verification. An address valid today can be closed tomorrow.
- Some servers lie by design. Catch-all servers accept everything, then may discard or bounce later.
- Full mailboxes and temporary outages make results fluctuate.
- Spam traps look valid. They are designed to be indistinguishable from real addresses; see spam traps.
A good verifier is honest about uncertainty: it labels unclear addresses as catch-all or unknown instead of pretending they are valid.
Deliverability guarantees
Some providers promise that addresses they mark as valid will not exceed a certain bounce rate, or they refund credits. DeBounce, for example, advertises a deliverability guarantee on its pricing page. These guarantees are useful signals of confidence, but they always come with conditions: the time between verification and sending, the types of addresses included, and how bounces are measured. Read them before relying on them.
How to test a verifier on your own data
The only accuracy figure that matters is the one measured on your list. Here is a fair method I use when comparing tools:
- Take a sample of about 1,000 addresses from your real list, mixing recent and old contacts.
- Verify the same sample with each tool, on the same day.
- Compare the results in a spreadsheet: where do the tools agree, where do they disagree?
- Send a normal campaign to addresses that all tools mark as valid, and note the bounce rate.
- Send carefully to a small batch of the disagreement group, and see which tool was right more often.
- Judge on bounces among "valid" results, the share of unknowns, and the price per useful result.
Sending to addresses marked invalid just to check the tool creates bounces that hurt your reputation. Test on the valid and uncertain groups only.
Reading your results
| What you observe | What it suggests |
|---|---|
| Very low bounces on valid results | Good accuracy on what the tool confirms |
| Many unknowns | Cautious tool, or difficult domains in your list |
| Few unknowns but more bounces | Tool may classify uncertain addresses as valid too easily |
| Large catch-all share | B2B audience; consider catch-all validation |
What to do with uncertain results is explained in "unknown" verification results.
What accuracy is worth
A slightly more accurate tool is worth more if your list is large, B2B-heavy, or if you are close to your platform's bounce limits. For a small consumer list cleaned twice a year, price, ease of use and credit terms often matter more than a small difference in accuracy on hard cases. Compare tools in best email verification tools.
Questions to ask a vendor about accuracy
- How do you define and measure accuracy?
- Which categories do you return for uncertain addresses (catch-all, unknown, risky)?
- Do you retry greylisted or slow servers, and for how long?
- Do you charge for unknown results?
- What are the conditions of your deliverability guarantee, if you offer one?
- How long after verification do you recommend sending?
Clear, specific answers are a good sign. Vague promises of "99%+ accuracy" without definitions are not.
Accuracy and timing
Verification results are a snapshot. The longer you wait between verifying and sending, the more the list changes. For an active consumer list, results stay useful for a few weeks; for B2B lists with high turnover, the window is shorter. Verify close to the campaign date, especially for old segments.
A simple way to track accuracy over time
| Campaign | Sent to "valid" only | Hard bounces | Bounce rate |
|---|---|---|---|
| January newsletter | 12,000 | 36 | 0.3% |
| February newsletter | 12,400 | 31 | 0.25% |
| March promotion | 12,900 | 52 | 0.4% |
The figures above are illustrative. If the bounce rate among "valid" results stays low and stable, your verification is working. A sudden rise usually means the list changed (an import, an old segment) or that too much time passed since verification.
Glossary of accuracy terms
| Term | Meaning |
|---|---|
| False positive (for validity) | An address marked valid that actually bounces |
| False negative | An address marked invalid that actually works, leading you to remove a real contact |
| Resolution rate | Share of addresses that get a definitive valid or invalid result |
| Bounce rate on valid | Share of "valid" results that bounce when sent; the most practical accuracy measure |
Both kinds of error have a cost. False positives create bounces; false negatives make you delete real customers. A good tool keeps both low and is honest about uncertainty in between.
What to do with the results of a comparison
If two tools perform similarly on your sample, choose the one with the better price, credit terms and workflow. If one clearly produces fewer bounces among valid results on your audience, that advantage is often worth paying for, especially on large or B2B lists. Either way, keep your bounce rate as the final judge after each campaign.
Summary
- Accuracy claims are not comparable across vendors
- Measure bounces among "valid" results on your own list
- Expect uncertainty on catch-all and protected domains
- Test two tools on the same sample before a large purchase
- Send soon after verifying
Keep results accurate over time
- Send soon after verifying; results age
- Re-verify segments not emailed for 3 months or more
- Remove hard bounces immediately after each campaign
- Verify new signups in real time so bad data never accumulates
- Track your bounce rate per campaign as the final accuracy check
Frequently asked questions
Can any tool be 100% accurate?
No. Some servers hide whether mailboxes exist, and mailboxes can close right after verification. Good tools are highly accurate on clear cases and honest about uncertain ones.
Why do two verifiers give different results?
They use different infrastructure, retry strategies and rules for catch-all and unknown results. Differences are mostly in risky categories.
What is a deliverability guarantee?
Some providers promise a maximum bounce rate on addresses they mark as valid, under conditions. DeBounce, for example, advertises a deliverability guarantee on its pricing page. Read the conditions before relying on it.
How do I test a verifier fairly?
Use the same sample, verify it with each tool on the same day, then send only to addresses all tools mark valid and to those where they disagree, in small batches, and compare bounces.
Does verification accuracy decline over time?
The results age. An address valid today can be closed next month, so verify again before sending to older segments.
