Who actually receives B2B email

Why the gateway number is the one that matters

Microsoft and Google filter, but they filter a general population of mail. A security gateway is different: it is a product a company bought on purpose to stop unsolicited email before it reaches a mailbox, and it is usually deployed by larger and more regulated organisations. 493,304 domains, 5.88 percent of the population, sit behind one.

That number is small and it is not evenly distributed. It concentrates in exactly the companies an enterprise-focused outbound programme most wants to reach, which is why a campaign aimed up-market can bounce far harder than its list quality alone would predict. Our own campaign work has run into this directly: gateway rejections, not invalid addresses, drove the bounce rate on programmes targeting larger organisations.

The practical consequence is not "avoid these companies". It is that a bounce rate is not a single number. Segment it by receiving provider before concluding that a list is bad, because a 1 percent bounce on Google-hosted domains and a 6 percent bounce on gateway-hosted domains are two different problems with two different fixes. The thresholds we run to are on deliverability, and what our own campaigns actually measured is on cold email benchmarks.

What this changes about sending

• Two rule sets cover most of the market. At 54.32 percent combined, authentication and reputation practice that satisfies Microsoft and Google covers the majority of any B2B list. That is a simplification worth having.
• Volume per mailbox is a Google and Microsoft question. Their acceptance behaviour is what most of your list will apply, so warmup and per-mailbox pacing should be tuned to them rather than to an average of everyone.
• The long tail is real but small individually. 30.89 percent of domains are self-hosted or on smaller and regional hosts, and no single host in that tail accounts for more than about 3 percent of it. There is no third giant to optimise for.
• Domain-level is not people-level. This census counts domains, not employees or leads. A large enterprise and a two-person consultancy are one row each.

The largest hosts inside the long tail

For completeness, since a bucket labelled "other" is the weakest thing a study can publish. These are the biggest individual MX hosts among the domains we could not attribute to a named provider.

Method

Every distinct email domain in a 188 million-lead B2B contact database was extracted and deduplicated, giving 9,847,723 unique domains. Each was resolved for MX records with aiodns over rotated public resolvers, in five passes so that transient timeouts converge rather than being recorded as failures. Resolution completed on 19 July 2026.

Of those, 8,384,502 returned an MX record and form the population for every share on this page. The remainder break down as 552,369 with no MX but an A record, which is still deliverable under RFC 5321, 662,748 that did not resolve at all, 154,499 with no records, and 93,605 still unresolved after five passes and therefore not claimed either way.

Providers were assigned by matching the MX hostname against a fixed, ordered rule list. Security gateways are matched before tenant hosts, so a Microsoft 365 tenant behind Mimecast is counted as Mimecast. That is deliberate: the gateway is the machine that accepts or refuses the connection. Domains whose MX host contains their own domain name are counted as self-hosted. Everything else is the long tail.

Limitations

• Domain-level, not mailbox-level: one domain is one row regardless of how many leads sit behind it, so this is not weighted by company size.
• Gateways are classified ahead of tenant hosts, so a Microsoft 365 tenant sitting behind Mimecast counts as Mimecast. That is deliberate: the gateway is the box that accepts the connection, and it is what decides whether the message arrives.
• 'Self-hosted (own domain)' means the MX host contains the domain's own name. That is a good proxy for self-hosting but it will also catch some hosts that white-label onto a customer subdomain.
• 'Smaller and regional hosts' is a long tail, not one population. The largest single host in it accounts for under 3 percent of that bucket.
• The population is the email domains present in one B2B lead database, not the internet. It skews to companies that appear in commercial contact data.
• Resolved on 2026-07-19 across 5 passes. MX records change.

The classification rules, the counts and the residual breakdown are all reproducible from the same resolution database, and every share on this page is a single division of two counts stated above.

Where to go next

For what our own campaigns measured, including a 14x spread in reply rate between campaigns, see cold email benchmarks. For the sending-side practice this implies, see deliverability. For what an outbound programme costs to run at volume, see what a cold email agency costs, and for who should own each part of it, done-for-you outbound.