Article

Do Email Verifiers Really Have 99% Accuracy? What the Number Means

By Unlimited Verifier Team ·

Diagram showing that an email verifier's 99% accuracy claim measures agreement with a test set, not the bounce rate of a real email list.

Summary

No. A 99% accuracy figure usually reflects how often a verifier matches a vendor-built test set, not how many of your emails will bounce. The metric that predicts bounces is the false-valid rate, which you can measure on your own list with a small sample of known outcomes.

No, not in the way most buyers read it. A 99% figure is usually a vendor's score on its own labelled sample, so it measures how often the verifier agrees with those labels, not how many of your emails will bounce. Unlimited Verifier states 99.5% for its deep verification, which is a claim to test rather than a number to accept. For the wider picture, see our guide to email verifier accuracy.

Quick answer

Where does the 99% accuracy claim come from?

Most 99% figures come from the vendor, using a sample the vendor assembled, and the full method is rarely published. That doesn't make the number false. It means you are reading a result, not a verified specification.

A typical test takes addresses whose status is already known, often from past bounce data or manual checks, runs them through the verifier, and counts matches. Three choices shape the headline before a single address is checked:

ZeroBounce and NeverBounce both publish accuracy claims, yet the same address can come back with different labels from each. Our explanation of why ZeroBounce and NeverBounce disagree on the same email covers the mechanics. Independent tests exist, but they are far fewer than vendor claims, and their results change with the sample and the date. For how to weigh them, see our independent accuracy benchmark and our guide to whether vendor accuracy claims are trustworthy. When you compare the best email verification tools on accuracy, judge each claim by its method, not its headline.

Bottom line: A 99% figure describes one test set, built under choices the vendor controls. Treat it as a claim to verify.

Why doesn't 99% accuracy mean 99% of your emails will be delivered?

Accuracy averages over both kinds of address, and most lists are lopsided. The number that predicts bounces is narrower: among addresses tagged Valid, how many actually bounce.

Take a 100,000-row list where 95,000 addresses are deliverable and 5,000 are not. A tool that tags everything Valid scores 95% accuracy and checks nothing. A tool at 99% accuracy makes 1,000 errors, and how those errors split decides what you experience:

The same 99% can mean zero bounces with 1,000 lost contacts, or a bounce rate that alone breaches your target. The headline cannot tell you which. Even a correct Valid result doesn't guarantee inbox delivery, a point we unpack in whether a Valid result means the email will be delivered.

Bottom line: Overall accuracy averages away the error that hurts you. Track the false-valid rate within the Valid segment.

What does a 99% accuracy claim leave out?

It usually omits the Risky and Unknown shares, the breakdown by provider, and the test date. The table below shows how to decode common phrasing.

Claim as written What it usually means Question to ask
"99% accurate" Agreement with the vendor's labelled sample Which sample, how many addresses, who labelled it?
"99% of emails verified" Share of addresses that got a decisive result What share came back Unknown or Risky?
"99% deliverable" The mailbox accepted the test, not that mail reached the inbox Does the figure predict bounces, complaints or inbox placement?
"Accurate across all providers" An average over mixed domains What are the results for Gmail, Outlook, Yahoo and corporate domains?
Third-party benchmark One reviewer's sample at one moment Sample size, collection date and method?

Two gaps matter most. Catch-all domains accept mail for any local part, so individual mailboxes on them can't be confirmed. Any figure that includes them needs a separate explanation, which our article on how accurate verifiers are on catch-all domains explores. Provider-level results also vary, so accuracy on Gmail versus Outlook addresses belongs on your vendor checklist.

Bottom line: An accuracy figure is only as informative as the segment it covers. Ask for the Unknown share and the provider breakdown before you believe it.

How big a test sample does a 99% claim need?

About 500 addresses gives roughly ±0.9 percentage points of sampling error at 99% accuracy. Below 200, the error band is wider than the gaps between most claims you will compare.

The 95% margin of error for a proportion is approximately 1.96 × √(p(1−p)/n). At p = 0.99:

Sample size (n) Margin of error (±) Range a 99% result could reflect
50 2.8 points 96.2%–100%
200 1.4 points 97.6%–100%
500 0.9 points 98.1%–99.9%
1,000 0.6 points 98.4%–99.6%
10,000 0.2 points 98.8%–99.2%

This covers sampling error only. It assumes a random sample with correct labels. If the ground truth is wrong for 2% of addresses, no sample size rescues the result. A large but biased sample has the same problem: 10,000 clean corporate addresses say little about a stale consumer list.

Bottom line: Around 500 well-chosen addresses with trustworthy labels is enough to see a gap of a few points between tools.

Which verification checks can actually prove an address is real?

Each layer proves something narrower than the word "valid" suggests. Most accuracy disputes arise when one layer is asked to prove more than it can.

Layer What it can establish Where it fails
Syntax The address is well-formed A correctly formatted address on a mistyped domain passes
Domain and MX The domain exists and publishes mail servers A live domain can still have no such mailbox
SMTP mailbox check The receiving server answers the recipient check Catch-all servers accept any local part, so the answer is ambiguous
Disposable and role flags The address is a throwaway, or a shared inbox like info@ They don't prove consent or engagement
Your engagement and bounce history Whether people open, click or reply Unavailable for new or imported contacts

The SMTP layer is the most powerful and the most misread. SMTP reply codes are defined in RFC 5321, and enhanced status codes such as 5.1.1 (bad destination mailbox) are defined in RFC 3463. Receiving servers are not obliged to answer the recipient check honestly, and some accept everything. So an Unknown result isn't necessarily a verifier failing. It often means the server gave no usable answer.

Bottom line: A verifier's accuracy depends on which layers it can run against each address. Mailbox-level proof is available only on some domains.

Which error costs more: a false valid or a false invalid?

For most senders, a false valid costs more in deliverability, and a false invalid costs more in lost revenue. Know both before you set a threshold.

Platform thresholds make the first error urgent. Amazon SES's documentation describes a 5% bounce rate as a review trigger and 10% as one that can lead to pausing. Gmail's sender guidelines warn against user-reported spam at or above 0.3% and advise staying below 0.1%. Our recommendation is stricter than either: keep total bounces under 2%, and keep bounces from the Valid segment under 1%.

Bottom line: Count false valids as bounces and false invalids as lost contacts. Set your threshold on the first.

How do you test a verifier's accuracy on your own list?

Build a ground-truth sample from your own sending history, run it through each verifier, and compare every tag with what happened. Your bounce logs describe your addresses, which makes them a better test set than any generic sample.

  1. Choose ground truth. Use addresses you mailed in the last 60–90 days, labelled with the ESP's delivery status. Hard bounces are "bad" and delivered addresses are "good". Wait several days before labelling, since late bounces can arrive after a server has accepted a message.
  2. Understand what "delivered" means. It usually means the receiving server accepted the message. It does not mean the message reached the inbox, so treat it as a proxy.
  3. Sample the messy part. Include older contacts, imported lists and catch-all domains. Testing only engaged customers gives the tool easy cases and inflates its accuracy.
  4. Target 400–500 addresses. That gives roughly ±1 point at 99% accuracy.
  5. Run identical input through each tool on the same day. Results can shift as mail server configurations change.
  6. Build a 2×3 table of Valid, Risky and Invalid against Delivered and Bounced.
  7. Compute four numbers: the false-valid rate (bounced ÷ Valid-tagged), the false-invalid rate (deliverable addresses tagged Invalid ÷ all deliverable addresses), the Risky split, and decided accuracy (Valid and Invalid results only).

Here is a hypothetical 500-address sample, shown to illustrate the arithmetic rather than to benchmark any tool:

Verifier tag Delivered Bounced Total
Valid 405 10 415
Risky 15 5 20
Invalid 5 60 65
Total 425 75 500

That sample falls short of 99%, and its 2.4% false-valid rate matters more than the headline. Sending 40,000 Valid addresses at that rate would produce about 960 bounces. Accuracy hides that; the false-valid rate shows it.

Bottom line: Test on your own ground truth. The false-valid rate is the number that predicts your bounces.

Hands-on: run the test on Unlimited Verifier's Free plan

This walkthrough uses the 500-address method above and stays within the Free plan's limits.

  1. Export 500 addresses from your ESP as a CSV. Keep the delivery status in a separate spreadsheet so the upload file contains only email addresses.
  2. Split the export into four files: 150, 150, 150 and 50 rows. The Free plan allows 25 lists a month of up to 150 emails each. Sign up for free if you don't have an account yet.
  3. Upload each file. CSV and Excel files up to 10MB are accepted, and the email column is detected automatically.
  4. Select deep verification. The 99.5% figure refers to that tier, so test on it. The Free plan includes 750 deep verifications a month, which covers the 500 addresses.
  5. Export results as CSV the same day. Free-plan results are kept for 24 hours only.
  6. Join results to your delivery-status column on the email address, using XLOOKUP or VLOOKUP.
  7. Fill in the 2×3 table and calculate the four numbers.
  8. Spot-check surprising addresses with the free email checker.

If the test shows a result you want to repeat each quarter, the email verification API on Pro includes 100,000 emails a month and accepts bulk batches of up to 10,000 emails per request. For how automated checks fit a workflow, see email verification API and automation.

Bottom line: A 500-address test fits in the Free plan and gives you a false-valid rate you can defend.

Is Unlimited Verifier's 99.5% figure reliable, and what does it apply to?

Unlimited Verifier states 99.5% accuracy for its deep verification. That is the company's own figure. This guide is based on public documentation, not an independent benchmark, so confirm it with the test above before relying on it.

Compare email verification pricing on cost per 1,000 verifications rather than headline plan price. On Pro, $29 a month for 25,000 deep verifications works out to $1.16 per 1,000 at full use. Billed annually at $25 a month, it is $1.00 per 1,000.

Bottom line: The 99.5% figure is a starting claim. Your own false-valid measurement is the proof.

How do you verify a 40,000-contact list within plan limits?

Split the list into files of 25,000 or fewer, run standard verification on everything, and spend deep credits on the riskiest addresses first. Consider an ecommerce brand with 40,000 contacts: 12,000 from a 2019 purchased import and 28,000 from its own checkout. Our guide to email verification for ecommerce and SaaS covers why imports and checkout lists decay differently.

  1. Split by the Pro per-list cap. Pro allows lists of up to 25,000 emails. List A holds the 12,000 imported contacts plus the 13,000 oldest checkout contacts, totalling 25,000. List B holds the remaining 15,000 checkout contacts (28,000 − 13,000 = 15,000).
  2. Run standard verification on all 40,000. Standard is unlimited and carries no per-email fee.
  3. Spend deep credits on the riskiest addresses. Month 1: deep-verify List A, using 25,000 credits, the full monthly allowance. Month 2: deep-verify List B, using 15,000 credits and leaving 10,000 spare for re-checks.
  4. Calculate cost. Two months on monthly billing is 2 × $29 = $58, or $1.45 per 1,000 addresses deep-verified (58 ÷ 40).
  5. Export within 30 days. Pro results are kept for 30 days, so export both lists before the window closes.

Hypothetical outcome for the full 40,000 (illustrative only): 33,000 Valid (82.5%), 4,400 Risky (11%) and 2,600 Invalid (6.5%). Those add to 40,000. If your own test shows a false-valid rate near 2.4%, the Valid segment would produce about 33,000 × 0.024 = 792 bounces. That is why the test should come before the full send.

For signups, real-time validation at signup stops new bad addresses entering the list, so the cleanup doesn't recur.

Bottom line: Split to fit the 25,000 cap, verify everything in standard, spend deep credits on the highest-risk segments first, and export before the 30-day window closes.

What should you do with Risky results?

Keep Risky addresses out of your first sends, measure them separately, and never count them as Valid when reporting accuracy. Risky results are where headline figures quietly hide their uncertainty.

Risky typically covers catch-all domains, disposable and role addresses, and mailboxes that couldn't be confirmed. In the hypothetical 500-address sample, 15 of 20 Risky addresses delivered (75%), compared with 97.6% of Valid-tagged addresses. Risky is neither safe nor dead, so it needs its own sending decision. Spot-check catch-all results with the catch-all email verification tool. For context on acceptable levels, see what a good unknown rate looks like.

Bottom line: Treat Risky as its own bucket, test it separately, and keep it out of the Valid figures you report.

What are the common mistakes when reading accuracy claims?

  1. Treating the headline as a bounce forecast. Overall accuracy says nothing about your Valid segment. Track the false-valid rate.
  2. Testing only on your best contacts. Engaged customers are easy cases. Include old and imported segments.
  3. Running a 50-address trial. At that size, a 99% result carries about ±2.8 points of error, too wide to separate tools.
  4. Folding Risky into Valid to raise the number. This inflates accuracy and hides the bounces that matter.
  5. Comparing tools on different files or on different days. Use identical input and run them together.
  6. Letting results expire. Free-plan results are kept for 24 hours and Pro results for 30 days. Export your evidence the same day.
  7. Treating Valid as consent. A working mailbox isn't an opted-in subscriber. Review email verification compliance and hygiene and the FTC's CAN-SPAM compliance guide.
  8. Verifying once and assuming the list stays clean. Addresses decay as people change jobs. Follow our email list maintenance playbook and re-test the false-valid rate every quarter.

Bottom line: Most bad decisions come from reading one number as if it described the whole list. Measure the segment that will receive your mail.

How we compared (and limitations)

This guide is based on public documentation: the SMTP reply-code RFCs, Amazon SES's published thresholds, Gmail's sender guidelines and the FTC's CAN-SPAM guide. We haven't run controlled tests against any vendor, and we haven't independently benchmarked Unlimited Verifier's 99.5% figure. Thresholds and platform behaviour change, so check current documentation before acting.

The 500-address sample, the 100,000-row example and the 40,000-contact scenario are hypothetical. They show the arithmetic and plan limits, not measured results.

For a side-by-side look at tools, see our tool roundup and how to choose an email verifier. To understand where accuracy numbers come from, read how email verifiers measure accuracy and how accurate email verification really is.

Bottom line: A 99% claim is a starting point. Measure the false-valid rate on your own list before trusting any tool, including this one.

Worked example: 99% accuracy vs. false-valid rate

Assume a 100,000-row list with 95,000 deliverable and 5,000 dead addresses. A verifier at 99% accuracy makes 1,000 errors. Two scenarios show why the split matters:

Scenario Errors Dead addresses tagged Valid Valid tags False-valid rate
A: all errors are dead addresses tagged Valid 1,000 1,000 96,000 1.04%
B: all errors are good addresses tagged Risky 1,000 0 94,000 0%

In Scenario A, about 1 in 96 Valid sends bounces, even though accuracy is 99%. In Scenario B, no Valid sends bounce, but about 1,000 good leads are held back as Risky. Both scenarios have identical accuracy, so only the false-valid rate and the cost of held-back leads reveal which tool fits your list.

Frequently asked questions

Is 99% email verification accuracy realistic?

It can be achieved on a specific test set, but it is rarely a guarantee for your list. Accuracy depends on the sample, how Unknown and Risky results are counted, and who set the labels. Test a tool on a sample of your own addresses with known outcomes to see its real performance.

What is a false-valid rate?

The false-valid rate is the share of addresses a verifier tags as Valid that still bounce. It is the most useful number for predicting bounces, because it measures the risk of the tags you act on. You calculate it by dividing bounced addresses tagged Valid by all addresses tagged Valid.

Why can a verifier with high accuracy still cause bounces?

Overall accuracy averages over both good and bad addresses. If most of your list is deliverable, a tool can score high by getting common cases right while still missing some dead addresses. Those misses, tagged Valid, are what cause bounces.

How many emails do I need to test a verifier?

About 500 addresses with known outcomes gives roughly plus or minus 0.9 percentage points of sampling error at 99% accuracy. That is usually enough to separate tools with meaningfully different performance. Larger samples narrow the margin but cost more time and credits.

Where do known outcomes for a test list come from?

Good sources include past hard-bounce records from your sending platform, addresses you have confirmed through double opt-in, and manual checks on a subset. Use addresses whose status you trust, and avoid using the verifier's own output as the ground truth.

Should Unknown and Risky results count as errors?

They should be reported separately rather than hidden. Excluding them from the denominator inflates accuracy. Track how many addresses fall into each category, then decide whether to send to Risky addresses based on your tolerance for bounces and your sender reputation.

Do different verifiers agree on the same address?

Not always. Different tools use different checks, data sources, and thresholds, so they can label the same address differently. This is why a single accuracy claim cannot be applied across vendors. Compare tools on the same test sample to see the difference.

Are independent accuracy tests more trustworthy than vendor claims?

Independent tests are generally more useful because the tester does not control the sample or the labels. Still, results change with the sample, the date, and the method, so read the methodology before trusting a headline number.

Does a 99% verifier score mean my sender reputation is safe?

No. Sender reputation depends on your bounce rate, complaint rate, sending patterns, and list quality, not only on verifier accuracy. Verification reduces bounces, but it does not replace good list hygiene and consent-based collection.

Cut Bounces and Protect Your Sender Reputation With Unlimited Email Verification

Every invalid address you send to hurts your sender reputation and your inbox placement. Verifying your list before each send means fewer bounces, healthier deliverability, and a predictable cost. With Unlimited Verifier, standard verification is unlimited on every plan, with no per-email fees.

Unlimited Verifier checks syntax, domain, MX records and the SMTP mailbox. It also detects catch-all domains and flags disposable and role-based addresses. Export a CSV from your email platform or CRM, upload and verify the list, then re-import the clean results. You can also automate checks with the email verification API.

Free Pro
Price $0 $29/month ($25/month billed annually)
Lists 25 a month, up to 150 emails each Unlimited, up to 25,000 emails each
Deep verifications 750 a month 25,000 a month
API access No Yes (100,000 emails a month)
Result retention 24 hours 30 days

Why teams choose Unlimited Verifier

Unlimited Verifier states 99.5% accuracy for deep verification.

Create a free account and verify your next list before you hit send.

Want to test a single address first? Try the free email checker.