15.3 Service-Level Agreements and Service Credits

Quick Answer: An SLA is Microsoft’s financially backed commitment for a service — for many Microsoft 365 workloads the current Online Services SLA still targets 99.9% monthly uptime. If Microsoft misses it, you may receive a service credit (commonly 25% / 50% / 100% of applicable fees at the <99.9% / <99% / <95% tiers). Credits are not automatic cash. You file a claim. They are not Secure Score and they are not your backup plan.

The April 30, 2025 outline asks you to describe service-level agreements (SLAs) including service credits. Fundamentals candidates over-fit this bullet by memorizing every Azure VM tier. Stay with the Microsoft 365 idea: Microsoft publishes a contract metric, measures Downtime the way that service defines it, and — if you claim on time — may credit fees.

What an SLA is (and is not)

The current contract is the Service Level Agreement for Microsoft Online Services, published on Microsoft’s licensing site (English edition dated August 10, 2026 at the time this chapter was written). It is part of the volume-licensing agreement. Official opening language: if Microsoft does not achieve and maintain the Service Levels, you may be eligible for a credit towards a portion of your monthly service fees.

Definitions that show up on the exam in plain English:

• Service Level — the metric Microsoft agrees to meet (usually a monthly Uptime Percentage).
• Downtime — defined per service. Exchange Online downtime is the time users cannot send or receive email with Outlook Web Access. Teams downtime is the time users cannot conduct instant messaging or initiate online meetings. SharePoint downtime is the time users cannot read or write a site collection they should be able to use. OneDrive downtime is the time users cannot view or edit files in their work or school library. Those definitions are narrower than “the app felt slow.”
• Incident — one event or a set of events that cause Downtime.
• Applicable Period — for Microsoft 365–style services, the calendar month in which the credit is owed.
• Service Credit — the percentage of Applicable Service Fees credited after Microsoft approves the claim.

An SLA is not:

• Microsoft Secure Score or Identity Secure Score (security-posture percentages).
• A guarantee that users will never see an error. Official Azure reliability guidance is useful here too: a 99.9% SLA is not “the service is always up,” and the definition of downtime decides what counts.
• Service health text. Microsoft’s SLA states that outage communications are not confirmation of a missed Service Level or credit eligibility.
• Your backup, retention, or disaster-recovery design. Microsoft’s service-continuity story (redundant datacenters, service restoration) is Microsoft’s recovery of the service. It does not replace your responsibility to protect customer content you delete, to keep a second copy of critical files, or to plan for client-side or ISP failures the SLA excludes.

The 99.9 percent figure — verified, with honest exceptions

For many Microsoft 365 services, the August 2026 SLA still publishes this credit table:

That table is the one attached to Exchange Online (Outlook on the web send/receive), Microsoft Teams (chat and online meetings), SharePoint Online, OneDrive, Office on the web, and Microsoft 365 Apps (reduced-functionality mode caused by an activation issue). That is the number MS-900 has historically wanted when the stem says “Microsoft 365 SLA” without naming a specialty workload.

Do not pretend every related service uses 99.9%:

Previews, free tiers, and some volume-licensing purchase methods have different or no fee-based credits. Open / Open Value style purchases may receive service time (days) instead of a fee percentage. The SLA is organized service by service, and Copilot-branded products do not all share one number: Microsoft 365 Copilot is sold as an add-on on top of Microsoft 365, and the workload tables above are what cover the Exchange, SharePoint, OneDrive, and Teams data it works over. If a stem names a service you cannot find in the current SLA edition, the correct fundamentals answer is that Microsoft publishes the metric per service — look it up rather than guessing a percentage. Memorize the 99.9% + claim + credit pattern, and remember that the contract names the metric.

How a service credit actually happens

Official Claims language is the exam’s process question.

1. You submit a claim to Microsoft customer support with enough detail to validate it: description, time and duration, affected resources, number and location of users, and errors observed.
2. Deadline for Microsoft 365–style services: Microsoft must receive the claim by the end of the calendar month after the incident month. Example from the SLA: an incident on February 15 must be claimed, with required information, by March 31. (Azure claims use a different 60-day clock — do not mix them.)
3. Microsoft evaluates the data in good faith and aims to process claims promptly, typically within 45 days of receipt.
4. If a credit is owed, Microsoft applies it to Applicable Service Fees. You may not unilaterally offset your bill. Credits are the sole and exclusive remedy for performance or availability issues under the SLA.
5. Credits apply only to the affected service (or resource/tier), not to every Microsoft 365 SKU you own. If you bought Exchange Online and SharePoint Online separately (not as a suite) and both had downtime, you may need two claims.

Credits are not automatic. A Service health incident, a status.cloud.microsoft banner, or a tweet from a Microsoft status account does not deposit money. Someone with authority still files the claim.

What the SLA refuses to cover

The Limitations section is where “Teams is slow at our building” usually dies. Credits do not apply to issues outside Microsoft’s reasonable control (including network or device failure outside Microsoft datacenters, at your site, or between your site and Microsoft), third-party software or inadequate bandwidth, your misconfiguration, preview/trial bits, failure to follow required configurations, throttling of abusive or over-quota behavior, or latency without actual unavailability (unless that service has an explicit performance SLA). Exchange Online also calls out tenant misconfiguration, extensions, send/receive limit problems, and ISP / on-premises failures as outside the Outlook on the web availability level.

That list is why section 15.4 teaches you to separate “Microsoft is down” from “it is just us.” The SLA measures Microsoft’s defined Downtime, not every unhappy Friday afternoon.

Realistic exam-style scenarios

Finance heard “99.9 percent.” They ask whether Microsoft 365 can never fail. Teach the financially backed monthly target and the definition of downtime. A minute when Outlook on the web cannot send or receive counts for Exchange Online. A minute when a local ISP drops HTTPS may not.

Incident in February, claim in April. The SLA example is unkind to late paperwork: a February 15 incident must be claimed by March 31. April is too late for that Microsoft 365-style claim.

Service health post appears. Leadership assumes a check is in the mail. Official SLA text says communications are not confirmation of credits. Someone still files the claim.

Secure Score is 42. A stakeholder calls that “our SLA.” Secure Score is a security recommendation score. It does not pay credits and it does not measure Exchange uptime.

Deleted SharePoint library. An SLA credit, if any, is a fee adjustment. It does not undelete files. Retention, Recycle Bin, and backups are your design.

Copilot versus Exchange. A stakeholder asks which uptime number “covers Copilot.” The accurate answer is that the Online Services SLA commits per named service, and the 99.9% tables above are the ones covering Exchange Online, Teams chat and meetings, SharePoint Online, and OneDrive — the workloads Copilot reads from. Microsoft gives some Copilot-branded products, such as Copilot Studio, their own separate table with different tiers. Do not recite an uptime percentage for a service you have not found in the current SLA edition. If the stem says “Microsoft 365 services” in the generic MS-900 sense, 99.9% is the number that matches Exchange, Teams chat/meetings, SharePoint, and OneDrive.

Exam traps

• SLA ≠ Secure Score. One is a contract metric; the other is a security posture score.
• 99.9% is the common Microsoft 365 uptime target, not a universal number for every related cloud service.
• Credits require a claim with details and a deadline.
• Credits are fee adjustments, not cash refunds you take to a bank, and not a right to withhold payment on your own.
• A Service health incident is not an approved claim.
• Customer-side network, ISP, and misconfiguration sit outside typical SLA coverage.
• Credits are not backup or DR.

Official sources

• Service Level Agreements (SLA) for Online Services — current Online Services SLA download (August 2026 English edition used for the 99.9% and credit figures in this section)
• Service health and continuity — incidents versus planned maintenance, continuity versus your own DR
• How to read a service-level agreement — why the downtime definition matters as much as the percentage
• Study guide for Exam MS-900 — SLAs including service credits