SLA (Service Level Agreement)

The SLA is the promise with teeth: the quantified commitments a provider makes about availability, performance and response — and what you get when they miss. Its craft lies less in the numbers than in the definitions, exclusions and remedies that decide what the numbers actually mean.

What It Is

A Service Level Agreement defines measurable service commitments and consequences for breaching them: uptime (typically 99.9% — 'three nines' — or better, measured monthly), response and resolution times by severity, performance targets (page loads, API latency), support hours, and notification duties for incidents. Critical craft elements: how uptime is measured (calendar minutes vs business hours; whether 'degraded but up' counts), exclusions (scheduled maintenance, customer-caused issues, force majeure — which can hollow out the headline number), and remedies — usually service credits against fees, occasionally termination rights for chronic breach. The SLA lives inside a broader service or master agreement alongside responsibilities both parties accept.

Why It Matters to You

An SLA is risk pricing, and reading one is an executive skill. A 99.9% commitment means up to ~43 minutes of downtime a month is free — is that acceptable for the process it supports? Service credits capped at 10–30% of monthly fees mean the provider's worst case is a discount, while your worst case is an operational failure — which is why chronic-breach termination rights matter more than the credit table. And an SLA is only as good as its measurement: insist on independent status reporting or audit rights, define what counts as an incident, and check who decides severity. The most expensive SLA is the one that reads well and measures nothing.

When You'll Meet It

At every SaaS or managed-service contract, and at renewal when you finally have data on how the provider actually performed. Negotiate measurement and remedies with the same seriousness as price.