Best for
- Teams arguing reliability by anecdote ('it feels flaky')
- Setting targets before a customer contract sets them for you
- Deciding when to ship features versus harden systems, with numbers
What you give it
- Your critical user journeys, current measurements (or access to make them), and your appetite for the reliability/speed trade
What you get back
- Per journey: the indicator (what exactly is measured, where), the objective (the target with its window), and the reasoning
- The error budget mechanics: what is left to spend, what happens at exhaustion — the policy written before it is needed
- Honest measurement: counted at the user's edge, with the gaps between measurement and experience named
How it works
- Starts from journeys, not services: users experience checkout, not the auth microservice.
- Defines indicators precisely: the exact fraction (good events over valid events), measured where users are, with the exclusions stated.
- Sets targets from measured reality and business need — aspiration priced in nines gets expensive fast, and it shows the price.
- Writes the error-budget policy while everyone is calm: the thresholds and their consequences, agreed before the deploy they will block.
Example
You: Define SLOs for our checkout and search before the enterprise contract negotiation.
Result: Checkout: availability measured as successful completions over attempts at the edge (99.5% over 28 days — current reality is 99.3%, so the target comes with its gap plan), latency as p95 under 2s. Search: 99% with a lighter latency bar (users retry searches; they abandon checkouts — the difference is in the reasoning). The budget policy: at 50% burn, risky deploys need review; at exhaustion, feature freezes until recovered — signed before the first argument, which arrived in week six and was settled by the policy in one meeting.
Limits — please read
- SLOs need measurement that exists; missing instrumentation is named as the prerequisite it is.
- Targets tighter than the dependencies beneath them are fiction; it checks the stack's arithmetic.
- The budget policy only works if leadership honours it on the bad week — it writes the policy; the honouring is cultural.