What Happens If We Fail?
An SLA (Service Level Agreement) is an external, legally binding contract between a service provider and its customers. It defines what the provider promises to deliver and what happens — usually financial penalties or service credits — if that promise is broken.
The SLA is the public-facing, commercial layer on top of your internal SLO. It answers the question: "What are we promising our customers, and what do they get if we fail?"
The Hierarchy: SLI → SLO → SLA
| Concept | Audience | Consequence of Missing | Typical Threshold |
|---|---|---|---|
| SLI | Engineers | Awareness | Raw measurement |
| SLO | Engineering team | Internal escalation, reliability work | Tighter (e.g., 99.9%) |
| SLA | Customers, legal | Financial penalties, contract breach | Looser (e.g., 99.5%) |
Because the consequences of breaking an SLA are severe — refunds, contract termination, reputational damage — your SLA target must always be looser than your internal SLO. If your internal goal (SLO) is 99.9%, your external contract (SLA) might only guarantee 99.5%. This gap is intentional: it gives your team time to detect a problem, respond, and recover before you breach the customer contract.
What SLAs Typically Cover
- Availability: "The service will be available 99.5% of the time per calendar month."
- Support response time: "Critical incidents will receive a response within 1 hour."
- Data durability: "Stored data will have 99.999999999% (eleven nines) durability."
- Remedies: "For each 0.1% below the SLA threshold, customers receive a 10% service credit."
A common interview question is: "Why should your SLA be lower than your SLO?" The answer has two parts. First, the SLA is a legal commitment — you need a safety buffer so that internal incidents don't automatically become contract breaches. Second, the SLO is your engineering target, and you want to exceed it consistently. If your SLO is 99.9% and your SLA is also 99.9%, any SLO miss immediately becomes a legal liability. The gap between SLO and SLA is deliberate risk management.