# Service-Level Agreement

A service-level agreement is a formal commitment between provider and customer that defines service expectations and consequences for shortfalls.

- Type: artifact
- Domain: Engineering & architecture
- Level: Intermediate
- Last reviewed: 2026-09-22
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/service-level-agreement/>)

## What it means

A service-level agreement, or SLA, documents externally meaningful service commitments. It usually defines covered services, performance measures, target levels, exclusions, measurement periods, claim procedures, responsibilities, and remedies such as service credits. Unlike an internal SLO, an SLA creates a customer-facing business or legal obligation.

## Why it matters

An SLA makes expectations and accountability explicit before a disruption creates disagreement. It helps customers assess operational fit and gives providers a boundary for promises. Clear definitions also reduce disputes about what counts as downtime, which telemetry governs, and how remedies are requested.

## When to use it

Use an SLA when supplying a service to customers or internal units that need a formal commitment. Draft it with product, engineering, support, commercial, and legal input. The promised level should account for architecture, dependencies, maintenance policies, and the provider’s ability to measure compliance consistently.

## How it works

1. Define the covered service, eligible customers, and responsibilities of each party.
2. Specify indicators, thresholds, measurement locations, and reporting windows.
3. State exclusions such as scheduled maintenance or customer-caused failures precisely.
4. Describe notification, escalation, evidence, claim, and remedy procedures.
5. Review the agreement when the service, risk, or customer need changes.

## A practical example

A hosted analytics product promises eligible enterprise customers 99.9% monthly API availability. The agreement defines an unavailable minute, excludes announced maintenance, explains how regional measurements are combined, and lists tiered credits. Internally, the team operates to a stricter SLO so it has room before breaching the commitment.

## Where it breaks

Credits rarely compensate for the full business impact of an outage. Broad exclusions or provider-only measurement can make an agreement unhelpful, while aggressive promises can drive disproportionate cost. An SLA does not itself improve architecture, incident response, support, or trust.

## Common mistakes

- Copying contractual language without verifying that telemetry can calculate it.
- Setting the internal objective equal to the external commitment, leaving no margin.
- Hiding consequential exclusions or complicated claim requirements.

## Sources

1. [Service Level Objectives](<https://sre.google/sre-book/service-level-objectives/>) — Google Site Reliability Engineering · Foundational book chapter
2. [Google Cloud Service Level Agreements](<https://cloud.google.com/terms/sla>) — Google Cloud · Provider SLA directory

## Related topics

- [Service-Level Indicator](<https://wiki.soroush.uk/topic/service-level-indicator/>)
- [Service-Level Objective](<https://wiki.soroush.uk/topic/service-level-objective/>)
- [Incident Management](<https://wiki.soroush.uk/topic/incident-management/>)
