SLOs that survive contact with quarterly planning.

The hardest part wasn’t the technology — it was the conversation about what "reliable" meant for the business. We ran a workshop per critical-path service: what does the customer feel when this breaks? what’s the latency they notice? what’s a tolerable error rate? The SLOs that came out of those workshops looked nothing like the marketing page.

Instrumentation was already mostly there in Prometheus and X-Ray. The framework added one Grafana dashboard per service showing burn rate, error budget consumed, and projected exhaustion. EventBridge alerted at the four-week burn-rate threshold so teams had a quarter to react.

The first quarterly planning under the new framework was the test. Two teams negotiated to spend their remaining error budget on a risky migration; one team proactively asked for an extra week on a release because their burn rate was already at 80%. The framework had become a real planning tool, not a slide deck.