Service Level Agreement

A service level agreement (SLA) is a document — often part of a larger contract — that defines the performance standards a provider commits to delivering a customer on an ongoing basis: how available a service must be, how quickly issues get responded to, and what happens when those commitments aren't met. It governs a continuing relationship, not a one-time piece of work.

Key components

• Specific, measurable metrics — uptime percentage, response time by severity level, resolution time, throughput, or whatever the service actually needs to guarantee, each stated as a number that can be checked against real data, not a general assurance.
• Measurement window — the period over which a metric is calculated (monthly uptime, quarterly average response time) and how it's measured, since the same raw numbers can look very different depending on the window and method used.
• Support tiers — response and resolution expectations that vary by issue severity (a total outage vs. a minor cosmetic bug), rather than one flat commitment for every kind of issue.
• Remedies and credits — what the customer is entitled to when a target is missed: typically a service credit or fee reduction, stated as a specific formula rather than left to negotiation after the fact.
• Exclusions — circumstances the metrics don't count against the provider, commonly planned/announced maintenance windows, force majeure events, or issues caused by the customer's own systems.

SLA vs. statement of work

These cover different things and are easy to conflate:

• Service level agreement — defines ongoing performance standards for a continuing service: how well and how reliably it will run, measured and enforced repeatedly over time.
• Statement of Work — defines the specific deliverables, scope, and timeline of a discrete project: what gets built or delivered, once, by when.

A vendor relationship can have both: a statement of work for the initial implementation project, and an SLA governing the ongoing service once it's live.

Common pitfalls

• Unmeasurable or vague metrics — a commitment to "high availability" or "prompt support" with no defined number gives neither party anything to check performance against, and turns every dispute into an argument about interpretation.
• No defined remedy for a missed target — an SLA that states metrics but not what happens when they're missed functions as a aspiration, not a contractual commitment.
• Metrics that don't reflect what customers actually care about — a provider hitting 99.9% uptime measured in a way that excludes the outage that mattered most to the customer wins on paper while losing the relationship in practice.
• Measurement window chosen to flatter the provider — averaging over a long enough period can hide a severe short outage that the customer experienced acutely.
• No process for disputing a measurement — if the provider is the sole source of the data used to judge its own performance, the customer has no way to challenge a number they believe is wrong.

A note on legal advice

This is general guidance on what SLAs typically contain and why, not legal advice. What's enforceable, what remedies are appropriate, and how disputes get resolved vary by jurisdiction and contract — always direct the reader to have an SLA drafted or reviewed by a qualified lawyer before signing.

Learn more

• Statement of Work for the deliverables-and-scope counterpart an SLA is often paired with but shouldn't be confused with.
• Contract Review for the broader review discipline an SLA should go through before signing.
• Non-Disclosure Agreement for another agreement type commonly negotiated in the same vendor relationship.
• Vendor Management for the ongoing practice of managing a provider relationship an SLA is meant to support.