Microsoft's 18-Hour Outlook Outage Moved Your Thursday Sends to Friday

Microsoft's 18-Hour Outlook Outage Moved Your Thursday Sends to Friday
Exchange Online held roughly 18 hours of mail, then released it in one Friday-morning surge.

Microsoft's Exchange Online outage ran close to 18 hours, from 17:35 UTC on November 12 to 11:20 UTC on November 13, 2026, and Microsoft extended its delivery retry window to 48 hours. For senders, that means Thursday's campaigns to Microsoft 365 inboxes mostly landed Friday morning in one backlog surge. Don't resend, and treat November 12-13 results as tainted.

Most coverage of this outage is about banks, airlines and earnings calls that slipped. Fair enough. But if you send B2B email, a big share of your list sits on Microsoft 365, and the more boring question is what happened to the campaign you scheduled for Thursday afternoon. Short version: it probably did arrive. Just late, in a pile, and at a moment when nobody was reading email for pleasure.

What actually broke (and for how long)

Per PPC Land's reconstruction using Kentik and Cloudflare network data, Exchange Online traffic fell to about 85% below expected starting around 17:35 UTC Thursday and began recovering at 11:20 UTC Friday. Once things came back, traffic ran about 19% above expected as queued mail and mailbox syncs caught up. Microsoft extended the email timeout to 48 hours around noon UTC on Friday, and by about 14:00 UTC said queues were "significantly reduced and approaching normal levels."

CNBC reported the problems started shortly before 1 p.m. ET Thursday and lasted more than 20 hours for some users, with Outlook, Teams, SharePoint and Copilot all hit. Microsoft's own explanation, quoted in ABC7's live updates, was that "a process on the affected servers responsible for establishing secure connections becomes overwhelmed." Microsoft also said it found no sign of malicious activity.

The email-delivery detail that matters most came from XDA's running coverage: some users could receive mail but could not send it, and Microsoft repeatedly warned about "delayed email delivery" and backlogs. The status page on day one listed 19 affected services and only said delivery "may be delayed." It never said whether external senders were being queued, deferred or rejected, which is annoying, because that's the one thing marketers needed to know.

Your Thursday send arrived Friday, probably in a pile

Here's how it most likely played out for a typical ESP send. Your ESP tried to hand mail to Microsoft's servers. Some of it went through and sat in Microsoft's queue, some got temporary failures (4xx deferrals), and your ESP kept retrying on its own schedule. Twilio SendGrid, for example, retries a deferred message for 72 hours from the first deferral. An 18-hour outage fits inside that window comfortably.

So delivery is probably fine. Engagement is the mess. Think of it like a restaurant that lost power during dinner service: the food still goes out once the lights come back, but every table gets served at once and nobody remembers which dish they ordered.

Your Thursday 2 p.m. ET newsletter competed on Friday morning with every other sender's Thursday newsletter, plus a day of internal mail people actually had to answer. I'd expect Microsoft-hosted recipients to show lower open and click rates on those sends, and I'd expect the drop to have nothing to do with your subject line.

The send didn't fail. The timing did, and your reports will blame the copy.

Do not hit resend (seriously)

The instinct when a campaign looks flat is to resend to non-openers. Don't, at least not this weekend. If Microsoft held the original for up to 48 hours and is still draining queues, a resend can arrive right next to the original. Two identical emails a few minutes apart is the fastest way to earn a spam complaint from someone who was perfectly happy with you last week.

Same logic for automations. This is the part I think most teams will miss. Any flow that branches on "did not open within 24 hours" between Thursday afternoon and Saturday probably sent the follow-up to people who couldn't open the first email. That includes onboarding nudges, webinar reminders, abandoned-demo sequences and sales cadences synced from your CRM. Pull the list of contacts who hit a non-open branch in that window and decide whether to pause the next step for them.

A 20-minute cleanup before Monday

None of this is hard. It's just the kind of thing that gets skipped because the outage is "Microsoft's problem."

  1. Tag the window. Mark every send between November 12, 17:30 UTC and November 13, 14:00 UTC as outage-affected in whatever you use for reporting. Exclude those sends from send-time tests, subject-line A/B results and any 2027 benchmark you build from Q4 data. A test that ran during this window is basically noise.
  2. Segment by mailbox provider, then compare. Most ESPs show deliverability or engagement by domain. Corporate domains on Microsoft 365 won't say "outlook.com," so look at your top B2B domains instead. If their open rate on the affected send is more than about 30% below their trailing four-send average while Gmail recipients look normal, that gap is the outage, not your list.
  3. Check soft bounces and suppressions. Some ESPs count each soft bounce toward an automatic cleanup. Mailchimp allows 7 soft bounces for addresses with no activity and up to 15 for engaged contacts before converting them to hard bounces. One outage won't hit those limits by itself, but if a contact was already close, it could tip them over. On SendGrid, a message that stays deferred past 72 hours puts the address on the Block suppression list. Check both lists on Monday for corporate addresses added November 12-16.
  4. Pause the "non-opener" branches. Covered above. Do it before the next automated step fires, not after.

If you already went through an Outlook deliverability check this month, this pairs well with it. Microsoft has also been quietly bouncing small senders without DMARC, and from what I've seen people are confusing the two problems. An outage bounce clears up on its own. A DMARC rejection does not, so if a domain still shows 550 5.7.1 errors next week, look at authentication rather than this outage.

The quieter lesson for anyone building on Microsoft data

One more thing, slightly off topic. PPC Land noted that Copilot and third-party AI assistants connected through Microsoft Graph stopped working when Exchange did, and that included connectors from Anthropic and OpenAI. If your team has started routing lead triage, reply detection or meeting booking through an AI agent that reads Microsoft 365 mailboxes, Thursday was a free stress test. Did anything alert you that replies weren't being processed? Or did you find out from a prospect?

And to be fair, none of this is new. Email has always been store-and-forward, and delays are built into the protocol. It just feels less forgiving now that so much of the funnel runs on automations that assume email arrives in seconds.

There's also a strategic angle, if you want one. Email has been the channel marketers lean on as organic search shrinks, mostly because you own the list. You own the list. You don't own the pipes, and a single provider going down for a day reminded everyone of that.

What I'd expect from Microsoft's post-incident report

Microsoft has promised a post-incident report covering root cause. My guess is it will confirm the overwhelmed connection process and a configuration change, and say almost nothing about what happened to inbound external mail during the window. I'd put the odds at maybe 70% that senders get no specific guidance at all, which means your own ESP logs are the only real record you'll have.

So pull them now, while the timestamps are fresh. Honestly, the teams I'd worry about aren't the ones whose Thursday campaign underperformed. It's the ones who'll put that number in a December deck and spend January trying to fix a subject line problem they never had.