In 2016, the modal business you cold-emailed ran its own mail server: 42.76% of Tranco top-1M domains with MX records pointed at self-hosted infrastructure, as our daily email infrastructure report shows across 192 DNS snapshots going back a decade. As of the 2026-07-05 snapshot, that figure is 22.53% — nearly halved. The domains did not disappear; they moved, mostly into Google Workspace (now 21.75%) and Microsoft 365 (16.68%). Most commentary treats this as a sad story about centralization. For anyone who sends outbound email for a living, it is the opposite: the single best thing that has happened to deliverability predictability in ten years.
Why self-hosted receivers were the hard mode
Every self-hosted mail server is a snowflake. The filtering stack is whatever the admin installed — SpamAssassin with decade-old score tweaks, a Postfix RBL list last reviewed in 2019, a greylisting daemon that delays your first message by fifteen minutes, or nothing at all. The practical consequences for a sender:
- No feedback loops. Gmail and Microsoft tell you, in aggregate, how you are doing. A self-hosted box silently drops you and never says why.
- Arbitrary blocklists. One admin's personal blocklist subscription can nuke an entire ESP IP range for their company, and no reputation-repair process exists.
- Inconsistent authentication handling. Some boxes ignore DMARC entirely; others run strict custom rules that reject anything unusual. You cannot optimize for both at once.
- Timeouts and greylisting that distort your sequencing tools' bounce and delivery metrics.
When a third of your list behaved this way, deliverability advice degenerated into folklore. Nobody could tell you what "the filter" wanted, because there were tens of thousands of filters.
The duopoly is a learnable system
Today Google and Microsoft receive mail for 38.4% of top-1M domains between them, and since 2022 Microsoft has been adding share 2.6 times faster than Google (+4.31 vs +1.64 percentage points). Two providers means two filter models, both documented, both stable, both with public sender guidelines:
- Gmail weighs engagement heavily — opens, replies, spam reports per sender — and publishes postmaster tooling to watch your standing.
- Microsoft 365 leans on reputation scoring and content analysis, with its own remediation and delisting processes.
A sender who learns these two systems now covers more than a third of any prospect list with known, optimizable behavior. Warm-up schedules, volume ramps and content patterns that work are reproducible across every Google-hosted or Microsoft-hosted prospect — something that was structurally impossible against a self-hosted majority. The segmentation play is covered in detail in our guide to segmenting outbound lists by mailbox provider.
Predictability cuts both ways. When two companies judge 38.4% of business email, a single algorithm change moves your numbers across the whole segment at once. The answer is monitoring, not nostalgia: watch placement per provider, and you will see shifts the day they happen.
The long tail did not get the memo
Before you delete your self-hosted playbook: 22.53% of domains still run their own mail, and another 12.45% resolve to mail hosts our 310+ classification patterns cannot identify at all — regional hosters, appliances, one-off setups. Together that is roughly a third of the top million still living outside the predictable duopoly. For this segment the old rules stand: conservative volume, plain-text leaning content, flawless authentication, and close bounce monitoring. We map that world in our long-tail deliverability guide.
What to do with this as an outbound team
- Split your list three ways — Google-hosted, Microsoft-hosted, everything else — with one MX lookup per domain. The first two segments get provider-tuned tactics; the third gets defensive ones.
- Expect the duopoly share of your list to keep growing. The self-hosted line has fallen in essentially every snapshot for a decade (42.76 → 30.18 by 2022 → 22.53 now). Every quarter, a slice of your "difficult" segment quietly migrates into filters you already understand.
- Reallocate effort accordingly. Time spent hand-nursing deliverability to individual self-hosted domains has a shrinking payoff; time spent mastering Gmail and Microsoft placement compounds.
The one discipline that covers all three segments is verification. Provider filters change without notice, and the long tail never told you anything in the first place. Running an inbox placement test before each campaign — seeing where your actual message lands at Google, Microsoft and a spread of smaller providers — turns the migration story above from an interesting chart into a working advantage.