Operations
SLA (Service Level Agreement)
A formal commitment to a customer about service performance (uptime, response time, resolution time) plus the remedies if you fail to meet it (typically service credits or termination rights).
In plain English
A promise you make in writing: 'we'll be up 99.9% of the time and respond to support tickets in under 4 hours.' If you break the promise, the customer gets a refund or other compensation.
Example
Enterprise SaaS SLA: 99.9% monthly uptime, P1 incidents acknowledged in 30 min, P1 resolution target 4 hours. Breach penalty: 10% monthly service credit per missed SLA tier. Total credits cannot exceed monthly fee.
Why it matters
SLAs are demanded by enterprise buyers and increasingly by mid-market. Without an SLA, you can't sell to procurement-driven companies. With a poorly-designed SLA, you can owe customers tens of thousands in service credits during a bad month. Most early-stage founders sign customer-written SLAs without modelling the exposure.
Common mistakes
- Accepting SLAs you can't measure — if you can't prove uptime, the customer's word becomes the metric
- Setting SLA tiers too aggressively — promising 99.99% with no engineering infrastructure to back it
- No cap on service credits — a customer could earn 200%+ of monthly fee in credits during a bad month
- Counting maintenance windows as uptime — must be excluded explicitly in the contract
- No carve-outs for force majeure (regional cloud outages outside your control)