The error budget is perhaps the most misunderstood concept in SRE. Most teams see it as a limit not to be crossed. High-performing teams see it as a resource to manage consciously — and spend wisely.
The formula
The error budget is the margin your SLO grants you over a given window. It is calculated directly from your objective.
For a 99.9% SLO over 30 days: (1 − 0.999) × 30 × 24 × 60 = 0.001 × 43,200 = 43.2 minutes. You have 43 minutes and 12 seconds of allowed outage this month. Not a minute more.
A resource, not a limit
An unspent budget is not success — it is a signal that you may be too conservative. Teams that never spend their budget tend to slow deployments, avoid changes, and let technical debt accumulate. A budget regularly consumed (without being exhausted) indicates a healthy deployment cadence and controlled reliability. The goal is not zero incidents — it is never exceeding your allowance.
The error budget policy
Any organisation serious about SLOs needs a written error budget policy — a short document that answers one precise question: what do we do when the budget is exhausted? The standard answer: no new features, focus on reliability until the budget resets. This policy gives SRE teams an objective lever to enforce a deployment freeze without negotiating case by case with every product team.
How burn rates affect the budget
The error budget is not consumed linearly. A full hour of complete outage burns 14.4 times more budget than a mild degradation spread over 30 days. This is exactly what burn rate measures: the speed at which your budget disappears, compared to the expected pace. A burn rate of 1 means you are on track. A burn rate of 14.4 means you will exhaust your monthly budget in under two hours.
Three common mistakes
- Using the budget to justify recurring planned maintenance — if your scheduled maintenance regularly exhausts the budget, your SLO is too strict for your operational model. Either revisit the objective or revisit the maintenance schedule.
- Measuring over windows that are too short — a 7-day window gives an error budget of 10.1 minutes for a 99.9% SLO. Any incident exhausts it instantly. Prefer 28 to 30 days for stable budgets.
- Having no error budget policy — without a pre-agreed rule for what happens when the budget runs out, every incident becomes a political negotiation rather than an operational decision.
Related articles
Try the simulator →
SLO Simulator →