Why we write shorter cold emails for Microsoft inboxes

We write a second, shorter version of every cold email for contacts whose company mail runs on Microsoft. Across five of our programs, the contacts on Microsoft replied far less often than the contacts on Google, and the blended numbers hid it. Splitting the results by mail host made the gap obvious.

Why do Microsoft-hosted contacts get their own email?

We looked at every contact we had emailed and sorted them by where their company mail lives. Microsoft-hosted contacts made up 4,046 of them and produced 9 human replies, which is 0.22 percent. Google-hosted contacts made up 1,286 and produced 23 human replies, which is 1.8 percent.

Auto-replies told the same story. Out-of-office replies came back at 2.1 percent from Microsoft-hosted contacts and 4.8 percent from Google-hosted ones. One program had 1,371 Microsoft-hosted contacts and zero human replies from any of them.

Microsoft-hosted contacts are also most of the list. They made up between 59 and 79 percent of the leads in most of those programs. One program, with domains registered in February 2026 and none blacklisted, showed no gap at all, so we left it alone.

Human reply rates differ by host for reasons besides delivery, since Google-hosted companies skew smaller. That is why we lean on the out-of-office rate, which should look similar for both hosts when mail lands everywhere. A seed test showed 4 of 4 Microsoft inboxes landing on the same domains, so we trust real sends over seed tests.

How do we tell which host a contact is on?

We read the public mail records for the contact's company domain. The MX record says which servers receive that company's mail, and a Microsoft-hosted domain points at outlook.com servers. We bucket each domain as Microsoft, Google, security gateway or other.

The MX record misses some companies. Some put a security gateway or their own mail server in front, and that server hands the mail to Microsoft 365 behind the scenes. On one program, 260 never-emailed leads sat in that group after our first split.

For those domains we check three more public records. We call a gateway domain Microsoft when its SPF record includes Microsoft's protection service, when its autodiscover name points at outlook.com, or when its DKIM records point at Microsoft. A Microsoft tenant lookup alone isn't enough, because it only says the company has some Microsoft account, not that its mailboxes live there.

How do we split a campaign?

Our sending tool can't aim a different email at part of a list inside one campaign. So we build a second campaign, which we call the twin, and move the right contacts into it. The twin copies the sender pool, the schedule, the daily caps and the settings, so length is the only thing that changes.

Only never-emailed Microsoft-hosted leads move. We stop each lead in the original campaign before adding it to the twin, so nobody gets both versions. Contacts we already emailed keep their original sender. For new programs we split by mail host on day one.

What changes in the short version?

The twin is the approved copy, shortened. Every email gets at most two short, plain sentences, aiming for about 25 words or fewer. Each sentence sits in its own paragraph, a plain statement first and then a short ask.

We tried one forced sentence first and our review rejected about 60 percent of those rewrites. Folding the hook and the ask into one sentence cost the email its taste and its proof, so two sentences won. We count words and not sentences, because two 45-word sentences aren't short.

When we shorten an email, we keep the sentence that carries the angle and cut the background. The offer can move to the second email. A short email that keeps only the background fact has lost its angle.

What stays the same?

• The offer, the claims and the call to action are the original's, and we add nothing new.
• The sender pool, the schedule, the caps and the sending settings are copied from the original.
• Campaigns built on custom opener and bridge hooks keep their copy as is, because flattening them breaks the personalization.
• Every email still gets spintax, about 6 to 12 spin points in the short version, and we run no A/B variants in a twin.
• The mailbox caps stay low for Microsoft contacts, which we cover below.

Do the mailbox limits change for Microsoft contacts?

Yes, and for the same reason. Microsoft filters are stricter, so a warmed mailbox sends up to 5 cold emails a day to Microsoft contacts against 15 for Google. A mailbox younger than 30 days sends 3 and 8, and we never run a mailbox at 30.

We also read the results by host in every campaign review. A healthy overall number can hide a Microsoft problem, as it did here. Our copy checker grades an email against the same rules, and the wider numbers sit on our cold email statistics page.

Common questions