How to Choose an Email Verification Service

Every email verification service on the market claims accuracy somewhere between 97% and 99.5%. We claim 98%. None of us can show you an independent third-party audit of that figure, and you have no practical way to check it before you buy.

So the accuracy number, the thing prominently displayed on every pricing page including ours, is close to useless as a purchasing criterion. Here is what is actually worth evaluating instead.

011. Does the free tier let you test the thing you are buying?

This is the criterion that resolves most of the others, because it converts marketing claims into observations.

The distinction to look for is not how many free credits you get. It is whether the free tier runs the same engine as the paid tiers, or a reduced version of it.

Several vendors give you generous free credits on a shallow check: syntax, domain, MX record. The mailbox-level scan, the part that determines whether a specific address actually exists, is paid-only. You test the free version, it seems fine, you pay, and the results are different because the pipeline is different.

Ask directly: does the free tier include the mailbox-level scan and catch-all detection? If the answer is no, the free tier tells you about the interface and nothing about the verdicts.

Ours is 500 credits a month on the full deep scan, the same engine the paid plans use. That is a deliberate choice and you should hold every vendor to the same question.

022. What happens to addresses the service cannot determine?

This is the criterion that separates careful services from confident ones, and it is the one most buyers never think to ask about.

Some addresses are genuinely undeterminable. A catch-all domain accepts mail to every address whether the mailbox exists or not, so acceptance proves nothing. A server that refuses the conversation tells you about its policy, not about the mailbox. These cases are real and they are not rare.

There are two ways to handle them, and they produce very different outcomes for you.

Guess, and report a status. The address gets marked valid or invalid based on a probability. Your reported accuracy goes up, because everything has a confident answer. Your actual outcomes get worse, because some of those guesses are wrong and you will send to them.

Report the uncertainty. The address is marked risky or unknown, and you decide what to do with it.

The second is less satisfying. Nobody enjoys paying for a verdict of "we could not determine this." It is also the honest answer, and it lets you make a decision that matches your risk tolerance. Sending to an unknown address is fine on an outbound campaign where a slightly higher bounce rate is acceptable. It is not fine when you are rebuilding a damaged sender reputation.

Ask what proportion of a typical list comes back in the uncertain buckets, and be suspicious of an answer near zero. Zero uncertainty on a real list means the uncertainty is being hidden rather than reported.

033. Can you see why, or only what?

When a service tells you an address is undeliverable, can you find out what led to that conclusion?

This matters in two specific situations, and if neither applies to you it matters less than the rest of this list.

The first is a dispute. A colleague insists an address is fine because they emailed the person last week. Without underlying evidence you have a disagreement between two assertions. With the recipient server's actual response you have a fact.

The second is a systematic problem. If a whole domain comes back undeliverable, you want to know whether the mailboxes genuinely do not exist or whether that provider is refusing checks from everyone. Those look identical in a status column and are completely different problems.

Most services give you a status and a confidence score. Some give you the underlying signals. If you regularly need to defend or debug results, this is worth paying for. If you upload a list once a quarter and trust the output, it is not.

044. How does the pricing behave when you are wrong about your volume?

Pricing models differ more than prices do, and the difference shows up when your estimate is off.

Pay per credit, credits expire. You buy 10,000, use 4,000, lose 6,000. Common and expensive if your volume is lumpy.

Pay per credit, credits do not expire. Better for irregular use. You buy once and draw down.

Flat monthly with an allowance. Predictable, wasteful if you underuse, and you need to know what happens at the ceiling. Does it stop, queue, or charge overage?

Per-address on demand. Fine for small volumes, expensive at scale.

The question that exposes the real cost is not "what is the price per thousand." It is "what happens if I buy the wrong amount." Ask about expiry, overage, and whether unused allowance rolls over, because that is where the money actually goes.

Ours is flat: $19 a month for 10,000 credits, $49 for 75,000, plus non-expiring top-up credits if your usage is irregular. Whether that beats per-credit pricing depends entirely on your volume pattern.

055. Does it fit the workflow you already have?

The most accurate verification service in the world is useless if getting data in and out is painful.

Things to check against your actual process:

CSV handling. Can you upload the file you have, with its extra columns intact, or do you have to rebuild it to match a template? Getting a cleaned file back that has lost the other twelve columns of your export is a real and irritating cost.

Where the results go. Do you export and re-import manually, or does it connect to your ESP or CRM?

API, if you need one. Check whether it is on your tier. It frequently is not on entry plans.

Speed at your volume. A 50,000-row list should not take a day. Ask about throughput, not just accuracy.

066. What does it do with your data after the job?

You are uploading personal data belonging to other people, and under GDPR, the DPDP Act and equivalent frameworks you carry responsibility for what happens to it.

Ask three concrete questions:

• How long are uploaded files retained, and are they deleted or just archived?
• Is your list used to train or enrich any shared dataset?
• Will they sign a data processing agreement if you need one?

The middle question is the one to press on. A verification service sees enormous volumes of address data, and there is commercial value in aggregating it. Some vendors are explicit that they do not. Some are quiet about it. Quiet is an answer.

07Questions worth asking any vendor

Copy these into an email and send them to your shortlist. The pattern of who answers plainly is informative on its own.

1. Does your free tier use the same verification engine as your paid plans, including mailbox-level checks?
2. On a typical list, what percentage of addresses come back as uncertain rather than valid or invalid?
3. Do you expose the underlying evidence for a verdict, or only the status?
4. Do unused credits expire?
5. What happens when I exceed my plan allowance?
6. How long do you retain uploaded lists, and do you use them for anything other than my job?
7. Will you sign a DPA?

08What no vendor can do

Two things worth knowing before you evaluate anyone, because they are sometimes sold as capabilities.

Nobody can reliably detect every spam trap. A trap is a normal mailbox that accepts mail. It looks exactly like a real address at the protocol level, because it is one. Verification reduces your exposure by removing the population traps hide inside, which is genuinely valuable and is not the same as detection. Any vendor implying otherwise is describing something that does not exist, and that includes us if we ever do.

Nobody can tell you whether someone wants your mail. A live address belonging to a person who forgot they subscribed verifies as deliverable, correctly. They will still hit the spam button. That signal is in your ESP's engagement data, not in a verification result.

09The short version

The accuracy percentages are unfalsifiable and all within a couple of points of each other. What genuinely differs between services is whether the free tier tests the real engine, whether uncertainty gets reported or buried, whether you can see the reasoning, how the pricing behaves when your volume is irregular, and what happens to your data afterwards.

Test on your own list before committing, because that is the only accuracy measurement that describes your data rather than someone's marketing page. We have side-by-side write-ups against ZeroBounce, NeverBounce, Bouncer, MillionVerifier and Clearout if you want our read, and you should verify it against your own results rather than taking ours.

Test on your own list, 500 credits a month, full scan, no card.