Outbound teams segment lists by industry, persona and company size, then send every segment through identical infrastructure with identical settings. The variable they skip is the one that decides whether the message is seen at all: which spam filter guards the recipient's mailbox. As of the 2026-07-05 snapshot of our daily email infrastructure report — a DNS scan of 664,715 Tranco top-1M domains with MX records — the receiving world splits roughly into four blocks: Google Workspace hosts 21.75% of domains, Microsoft 365 hosts 16.68%, self-hosted servers cover 22.53%, and 12.45% resolve to mail hosts no classification pattern matches. Different judges, different rules, different optimal plays.
Step 1: batch the MX lookups
You do not need an enrichment vendor for this. MX records are public DNS. For a one-off check: dig MX prospect.com +short. For a list, loop it:
while read d; do echo "$d $(dig MX $d +short | sort -n | head -1)"; done < domains.txt
Sorting numerically and taking the first line gives you the primary MX — the lowest-preference host, which is how our own classification works. Then map hostnames to segments: google.com / googlemail.com suffixes mean Google; mail.protection.outlook.com means Microsoft 365; an MX on the company's own domain suggests self-hosted; gateway names like Proofpoint or Mimecast mean a security layer sits in front; anything you cannot place goes to the long-tail bucket. A thousand-row list takes minutes.
Step 2: treat each segment as its own campaign
| Segment | Share of top-1M | What judges you | Tactical emphasis |
|---|---|---|---|
| Google Workspace | 21.75% | Engagement-driven filtering; Promotions tab | Warmed domains, conversational plain-looking mail, prune non-responders fast |
| Microsoft 365 | 16.68% | Reputation scoring and content analysis | Impeccable IP/domain reputation, conservative links and formatting, steady volume |
| Self-hosted | 22.53% | Whatever the admin installed — unpredictable | Plain text, perfect SPF/DKIM/DMARC, low volume, watch bounces closely |
| Unknown / long tail | 12.45% | Regional hosters, appliances, one-offs | Most conservative settings; expect silence instead of feedback |
Google-hosted prospects
Gmail's model rewards mail that people open and answer, and punishes mail they ignore. This makes list hygiene a deliverability lever: continuing to send step five to Google-hosted prospects who never opened steps one through four actively damages your standing with the whole segment. Cut non-engagers earlier here than anywhere else.
Microsoft-hosted prospects
Microsoft leans harder on sender reputation and content signals, and it is the segment that punishes infrastructure sloppiness most reliably. Sudden volume spikes, link shorteners and heavy HTML hurt disproportionately. This segment is also growing fastest — since 2022 Microsoft 365 has gained share 2.6× faster than Google, so whatever you learn here compounds. The playbook differences are detailed in our duopoly deep dive.
Self-hosted and unknown prospects
Together these two buckets are about a third of top domains — and the place where generic deliverability advice quietly stops applying. No feedback loops, no postmaster consoles, filtering that ranges from nonexistent to draconian. Default to the defensive posture: minimal formatting, no tracking-heavy templates, authentication that passes strict checks, and volumes far below what Gmail tolerates.
A message tuned for Gmail engagement can be exactly what a reputation-focused Microsoft filter or a rule-based self-hosted setup junks. When placement diverges per provider, observed reply rates diverge by multiples between segments even when the copy and the ICP are identical — you are measuring filter fit, not message quality. Segmenting lets you see and fix that per filter, instead of averaging it into one misleading campaign metric.
Step 3: operationalize it
- Add a provider column to your list at import time, from the batched MX pass above.
- Split sequences per segment — at minimum Google / Microsoft / everything-else — so volume ramps, content style and cut-off rules can differ.
- Report reply and bounce rates per segment, not per campaign. A blended number hides a dead segment behind a healthy one.
- Re-check MX quarterly. Domains migrate constantly — the decade trend is self-hosted shrinking into the duopoly, so your segment mix shifts under your feet.
The final piece is per-segment verification. Segment-level tactics rest on knowing where your message actually lands with each provider — and sequencer "delivered" stats cannot tell you inbox from spam folder. An inbox placement test across seed mailboxes at Google, Microsoft and a spread of smaller providers gives you that ground truth per segment before the campaign spends your domain's reputation finding out the hard way.