
Microsoft will not refund anyone for the Nov. 12 outage on its own. Its contract pays a service credit instead: money off a later bill for the service that failed. Customers have to ask for it.
We read Microsoft’s Service Level Agreement for Online Services, the document that decides these claims. The current edition is dated Nov. 17, 2026. It sets a 99.9% monthly uptime bar for each service separately. It also sets a hard deadline: for a November incident, Microsoft must have the claim by Dec. 31.
The SLA rides on a volume license, so home plans are out
The first line of the document does most of the work. This SLA “is a part of your Microsoft volume licensing agreement,” it says. That is the commercial contract behind business, enterprise, education and government subscriptions.
Buy Microsoft 365 Personal or Family and no such agreement exists. Consumer plans carry no uptime promise and no credit.
The SLA also shuts out anything free. “Previews and Online Services and/or service tiers provided free of charge are not included or eligible for SLA claims or credits,” it says. Trials and preview features are out too.
One group gets paid differently. Microsoft credits Open, Open Value and Open Value Subscription customers in service time, in days, rather than in fees.
Microsoft counts downtime per service, and it counts narrowly
Each service has its own definition of downtime, and the Exchange Online one is tighter than most people assume. It covers “any period of time when users are unable to send or receive email with Outlook Web Access.” Outlook on the desktop failing by itself does not count.
Teams downtime means users cannot hold instant messaging conversations or start online meetings. SharePoint Online means they cannot read or write part of a site collection. OneDrive for work or school means they cannot view or edit their files.
Microsoft 365 Copilot is the odd one out. The SLA does cover it, contrary to a claim doing the rounds. It counts unsuccessful responses from the model to valid prompts, and its first credit band starts below 99.5%.
Microsoft then measures uptime in user-minutes: each incident’s length times the number of users it hit. It weighs that against the total user-minutes in the calendar month.
Forty-three minutes buys the 25% band. Seven hours buys 50%
November has 43,200 minutes. Running Microsoft’s own formula against that gives the thresholds below, for a tenant whose entire user base lost the service.
| Monthly uptime | Service credit | Downtime needed in November |
|---|---|---|
| Below 99.9% | 25% | More than 43.2 minutes |
| Below 99% | 50% | More than 7 hours 12 minutes |
| Below 95% | 100% | More than 36 hours |
The outage ran from Microsoft’s first support case at 16:33 UTC on Nov. 12 to its final customer update at 04:15 UTC on Nov. 14. That is 2,142 minutes. The 100% band needs 2,160. By our arithmetic, the whole incident fell 18 minutes short of a full refund of the month’s fee for the affected service.
That is the outer envelope, not what any one tenant collects. Credits apply “only with respect to the actual downtime and not the overall outage duration,” the SLA says. Most tenants lost service for a fraction of that window. Anything past 7 hours 12 minutes still reaches the 50% band.
The admin center route, the incident IDs and Dec. 31
Microsoft takes claims through a support request. Admins sign in at admin.cloud.microsoft, choose Help & support, describe the problem and then select Contact Support, its guidance says.
The SLA lists what a claim must carry. It wants a detailed description of the incident, plus the time and duration of downtime. It also wants affected resource names, the number and locations of affected users, and a description of the errors. Claims that leave any of it out get rejected, the SLA says.
Quote the incident IDs. Exchange Online ran as EX1550543, the wider Microsoft 365 incident as MO1550821, and the earlier Teams fault as TM1550488. Admins can pull them from the Service health page in the admin center.
Microsoft must receive the claim by the end of the calendar month after the one in which the incident happened. The SLA gives its own worked example: an incident on Feb. 15 falls due by March 31. For Nov. 12, that is Dec. 31. Microsoft then aims to decide within 45 days.
If you bought through a reseller, you cannot file at all
Microsoft’s support page is blunt about this. “If you bought all your subscriptions through a partner, contact your partner for support,” it says.
Partners face their own rule. “Credit requests are accepted only from CSP direct and indirect providers,” Microsoft’s Partner Center guidance says. “Requests aren’t accepted from indirect resellers.” A small reseller therefore has to push the claim up to its distributor.
The provider needs the customer tenant’s GUID and the outage incident ID. It also needs proof the customer suffered the outage and asked for a credit. That proof has to come from an email address on the affected tenant’s domain, not a personal one. Non-Azure claims are due within one month of the end of the billing month.
A credit is not compensation, and Microsoft says so twice
Service credits are “your sole and exclusive remedy” for availability problems, the SLA says. They do not cover “lost revenue, operational costs, or any indirect losses.” A credit also cannot exceed that month’s fee for the affected service. The SLA allows one credit per service per month.
Microsoft adds a warning that will catch some claims. Its outage posts “are not a confirmation of missed Service Levels or Service Credits eligibility,” and advisory incidents are not usually eligible at all. Proving your own users were down is the claimant’s job. It is the same lesson as who actually collects a settlement.
The outage itself is over. Services recovered on Nov. 14, as we reported when the outage eased. Microsoft has since blamed a latent defect from a July update, a Teams failover and a 14% capacity shortfall. A preliminary post-incident report reached admins on Nov. 15, and Microsoft says a final review follows this week. Wait for that document before filing: it is the strongest evidence a claim can carry, and it arrives well before Dec. 31.


