The Error Budget Concept
What you'll learn: How error budgets transform the tension between shipping features and maintaining stability into a quantifiable, data-driven decision-making framework.
The Core Idea
An error budget is the amount of unreliability your service can tolerate before violating its SLO (Service Level Objective). It's calculated as the inverse of your reliability target: if your SLO promises 99.9% availability, your error budget is 0.1%—or about 43 minutes of downtime per month.
Think of it as a budget you can "spend" on unavoidability: outages, deployments, experiments, or technical debt. When your service is running well and exceeds the SLO, you have budget to spend. When incidents consume budget faster than expected, you must slow down and prioritize stability.
How It Guides Decisions
Error budgets create objective decision rules:
- Budget remaining? Green light for risky deployments, experiments, or feature velocity
- Budget depleted? Freeze feature work, focus on reliability improvements, implement stricter change controls
- Budget healthy but features stalled? You're over-investing in reliability; accelerate innovation
This framework resolves the classic conflict between product and engineering teams. Product wants features; engineering wants stability. The error budget provides a shared currency: both sides agree on the SLO, and the remaining budget becomes the neutral arbiter.
Example
Your SLO: 99.5% availability (217 minutes/month downtime budget). Current month: 120 minutes used in week 1 due to an incident. You have 97 minutes left for three weeks. The error budget signals: pause risky releases, prioritize the incident postmortem's action items, potentially defer the experimental feature until next month when the budget resets.
Key Takeaway: Error budgets quantify acceptable unreliability as a "currency" that balances feature velocity against stability risk, replacing subjective arguments with objective data about whether to ship or stabilize.