Friday the 13th & the Microsoft 365 Outage: What Happened and What it Means
Microsoft 365 had a rough Friday the 13th. Strictly speaking, the trouble started Thursday and ran well into Friday. As I write this, services are recovering, but Microsoft hasn’t published a root cause yet. Their post-incident reports tend to be very good, so we will know “soonish” what happened, but it raises some serious questions for anyone responsible for keeping an organization running on SaaS applications.
What Happened
Reports of Outlook problems began climbing around 17:25 UTC on Thursday, Nov. 12 (and Downdetector complaints peaked above 130,000. Exchange Online took the brunt as users were unable to open mailboxes, and mail was delayed or stuck. Microsoft later listed Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Planner, To Do, Intune, and Defender for Office 365 as potentially affected.
Microsoft’s explanation so far is narrow, however it seems a process on the affected servers that establishes secure connections might have become overwhelmed, causing the servers to stop processing requests. The company says the load didn’t come from a spike in traffic and that it has found no evidence of malicious activity. It also rolled back a recent network change, but hasn’t said whether that change was to blame.
Either way, recovery was bumpy, early fixes didn’t hold, and at one point servers Microsoft had restored failed yet again. More than 12 hours in, the company described only “marginal success.” By Friday afternoon in the US, it said email queues were approaching normal levels but I haven’t seen confirmation of full restoration as of this writing.
The Impact
Microsoft hasn’t said how many users were affected. Press reports put the reach at 74 countries, and Downdetector counts only the people who bother to report a problem. X, Threads, and LinkedIn were filled with reports and comments about outages and Forbes and other media reported disruption at businesses, airlines, and government agencies. Throughout the Pacific Rim, workers started Friday without Outlook, Copilot, or some Teams services, and various national and regulatory bodies took note.
The blast radius reached past Microsoft’s own apps. OpenAI said new connections to Microsoft services might fail, and Anthropic reported that email, calendar, and search tools in its Microsoft 365 connector were failing for most requests. Tools meant to protect and govern data were on the list, too. Microsoft’s affected-services list included Purview tools for data loss prevention and communication compliance, and some of them may not have worked properly while staff improvised.
So What, Right?
It seems like having a holiday from normal email would be welcome to lots of folks, but outages like this are worthy of concern. As someone whose career focuses on SaaS application resilience, with a long history in Microsoft-land, there are some interesting points here worth considering:
- “No data loss” doesn’t mean “no impact:” Microsoft said that this was an availability failure, not deletion or corruption of data. Having a backup doesn’t make Microsoft’s servers come back faster. But the incident raises the question we exist to answer “When our SaaS application platform is unreachable, can we still get to our own data?”
- Same-platform continuity plans aren’t real plans: Telling staff to move from Outlook to Teams can work until they fail together like they did in this case. Suffice it to say, we’ll all be surprised when we learn how many people ended up using WhatsApp, iMessage, or Gmail as side channels when their primary platform wasn’t available. Running email, chat, identity, and security tooling with one vendor or on a single cloud platform can be super-efficient… until it isn’t. Keeping an independent copy of data in a vendor-neutral cloud, on separate infrastructure, means the organization can still reach it if it loses access to Microsoft. Sales team can still find last quarter’s contract while mail is down, and it can be the difference between an attacker being able to pivot and hold at risk backups or holding them safe.
- Design for the first hour, not the full restore: In an outage like this, people generally don’t need a complete restore. They need the few things that keep work moving things like a customer’s contact details, a recent thread, or an on-call roster. That’s the lens that needs to be considered so ask questions like how fast can someone get usable access to the data that matters most, without depending on the thing that just failed?
- Response tools shouldn’t share the failure: Defender and Intune were on the affected list. So was Microsoft’s own service health tooling. When the tools the business uses to manage an incident degrade alongside the incident, a runbook stored in the same tenant becomes a liability so print it out and keep copies locally.
3 Questions to Ask This Week and Going Forward
- Which business processes stop if Exchange Online and Teams are both unavailable for 24 hours?
- Where does our incident runbook live, and can we open it without Microsoft?
- Do we hold an independent copy of the data we’d need, and have we tested getting to it?
Microsoft will likely publish a post-incident review, and we all will be reading it closely. Until then, the lesson doesn’t depend on the root cause. Cloud platforms will always have bad days, but what matters is whether you have a plan for them.









