Three-tier error budgets

Define error budget thresholds in three tiers—green (within budget, no action), yellow (approaching limits, investigate), red (exceeded, halt and redesign)—with pre-committed responses for each zone.

When This Fires

When implementing error budgets for any system (Define your error budget in writing, Every active goal needs an error budget — define…)
When a binary budget (fine/failed) doesn't provide enough early warning
When you want graduated response rather than binary reaction to budget consumption
When designing monitoring dashboards for personal or team systems

Common Failure Mode

Ignoring the yellow zone because "we haven't exceeded the budget yet": the budget is at 80% consumption with 60% of the period remaining, but since it's not red yet, no action is taken. This defeats the purpose of the yellow zone — it exists precisely to trigger investigation before the budget is fully exhausted. Yellow means "pay attention now so red doesn't happen."

The Protocol

(1) For each error budget (Express error budgets as numbers with time windows — '2…), define three thresholds: Green: 0% to X% of budget consumed. (X is typically 50-60% of budget at the period midpoint.) Yellow: X% to 100% of budget consumed, or consumption rate exceeding the sustainable pace. Red: 100%+ of budget consumed within the period. (2) Pre-commit responses: Green → no action, continue. Yellow → investigate: what's causing accelerated consumption? Can the trend be reversed? Red → halt and redesign: stop normal operation, conduct root cause analysis (When error budget is exhausted, analyze the pattern…), implement structural fix before resuming. (3) Check budget status at regular intervals — weekly for weekly budgets, daily for daily budgets. The tier should be visible without effort. (4) Never downgrade a red situation to yellow without structural intervention. "Things seem better this week" is not a fix.