We bounced 1.54% of 49,313 cold emails. Here's the list check behind it.

What this means for you

The numbers that matter

For context: Hunter's State of Email Outreach 2026, based on 31 million emails sent by Hunter users in 2025, says "The average bounce rate is 3.6%." Instantly's 2026 benchmark report tells senders to "Keep bounce rates below 2% (ideally much lower)." Different senders and different lists, so read that as context rather than a contest. The full verdicts, the split by provider and the bounce tables follow. If you want every public benchmark side by side, I put them in every public cold email benchmark, compared.

The full verdicts

Most of a B2B list is deliverable, and the part that isn't is small but expensive. Our verifier called 70.8% of 30,995 addresses safe, 22.6% risky, 4.8% invalid and 1.8% unknown. On what we actually sent, 757 of 49,313 cold emails bounced: 1.54%.

One test worth knowing about: 15,556 of these addresses came from a large data supplier that had already labelled them verified. Our own check still found 4.1% of them invalid and 15.2% risky. A supplier's "verified" is a claim about the day they checked, not about today.

Verdicts by mailbox provider

Provider from each domain's MX record, looked up on 20 October 2026.

You'll often read that Microsoft mailboxes always come back unknown, so "unknown" just means Outlook. In this B2B data that isn't true. Company domains on Microsoft 365 verified safe 78.8% of the time and came back unknown only 1.9%. Unknown was most common behind security gateways that sit in front of the real provider (9.9%). Consumer mailboxes like Hotmail and Yahoo may well behave differently, but only 16 addresses here are on one, which is too few to say. 71 addresses whose MX lookup failed or that sit on consumer domains are left out of the table.

The row that matters most is the last one. 267 addresses are on domains that have no MX record today, and 45.7% of them already verified invalid. A domain that stopped taking mail is usually a company that stopped, merged or renamed. That's decay you can see without a verifier.

Catch-all domains

6,634 of 27,859 company domains (23.8%) returned risky for at least one address. Risky mostly means a catch-all domain: the server says yes to every address, so the verifier can't tell a real mailbox from a made-up one. Role addresses and full inboxes also land there, and we don't store which reason applied. Risky was more common on Google Workspace (33.6% risky) than on Microsoft 365 (16.8%).

What happened when we sent to them, per address mailed:

Risky addresses bounced a bit more than safe ones, 3.3% against 2.1%. Invalid ones bounced 23.7% of the time. Some of those were mailed before the invalid verdict was recorded: a verdict here is the latest one on file, not necessarily the one at send time.

By provider of the address mailed:

Decay over time

I can't measure it yet, and I'd rather say so than make a number up. Measuring decay needs the same address checked twice, months apart. Our records keep the latest verdict only, and the verification date is set on just 2,738 rows, all from September 2026. So there are no repeat checks to compare.

The figure the market quotes comes from HubSpot's database decay simulation, which says "Marketing Sherpas research shows that B2B data decays at a rate of 2.1% per month", annualised to 22.5%. That page is a simulation built on a research figure, so treat it as a model. Cognism's data decay article repeats the same 22.5% and credits it to HubSpot, which is how one number turns into "the industry figure". The calculator below starts from that model because it's the only widely used one; change it if you have your own.

List decay calculator

Formula: list size × (1 − (1 − monthly rate)^months). Decay compounds: each month takes its share of what's still good. This estimates addresses gone bad since the last check, on top of whatever was already invalid then. Runs in your browser, nothing is sent anywhere.

What I do before every send

• Verify every export. Whatever the supplier says. The 4.1% above is why.
• Drop only invalid and disposable. Those are provably dead. Everything else stays.
• Keep risky and unknown. Risky is mostly catch-all corporate domains, and dropping them throws away real buyers to save a few bounces. Worth saying: the 78 unknown addresses we did mail bounced 35.9% of the time. That's a small group, and it cuts against this rule, so it's on the page rather than left out.
• Never delete a contacted lead. Someone who got an email, or replied, stays on record, even if a later check fails. Deleting them is how you mail the same person twice.

If you're setting up the list side from scratch, I wrote up how the whole prospecting pipeline fits together in building an AI prospecting system for B2B, and the cold email domain checker covers the sending side.

Method

• Verifier. A self-hosted SMTP verifier (Reacher). It connects to the receiving mail server and asks whether the mailbox exists, without sending anything.
• Addresses. 59,445 lead records as of 20 October 2026; 30,995 carry one of the verifier's four verdicts.
• Vocabulary. The status column mixes our verifier's verdicts with supplier labels and outreach states. Only the four verdicts are counted:

• Provider. From each domain's MX record, looked up once on 20 October 2026. A security gateway hides the real provider, so it gets its own row.
• Bounces. 49,313 cold emails sent 29 December 2025 to 20 October 2026 to 16,519 addresses, as reported by the sending platform. Win-back mail to a company's own former users is left out.

Limitations

• A few sources and ICPs. Mostly European B2B lists from a handful of suppliers, built for outreach. Not a random sample of all business email.
• Survivor bias. When an address is mailed, its status is overwritten with "sent" or "replied" (10,313 records). Those had passed the check before they were mailed, so the shares above under-count the good addresses.
• SMTP can't see every mailbox. Catch-all domains and servers that refuse to answer hide the truth from any verifier.
• Bounce data only for what was sent. Nobody mails the invalid ones on purpose, so the invalid bounce rate rests on 329 addresses.
• Today's MX, not then. Providers are classified by today's DNS, which may differ from the day of the check.