The Incident Feedback Loop That Compounds

Each incident teaches. The loop that captures lessons and applies them.

Capture

Capture happens at postmortem time. Each incident produces a contributing-factors list and an action-item set; cross-incident patterns surface only from the aggregate. Structured tags per incident support the pattern queries that come later; named quarterly pattern lead catches the "we never actually looked across incidents" failure mode.

• Postmortem identifies factors and items. Contributing-factors list and action-items set per incident. Standard postmortem output.
• Cross-incident pattern review. Per-quarter pattern review across postmortems. Five incidents with the same factor signal a systemic issue.
• Structured cause tags. Cause-class tags per incident. Supports later pattern queries against the archive.
• Named pattern lead per quarter. Responsible analyst per quarterly review. Catches "we never actually looked across incidents."

Apply

Application is what makes capture pay back. Items ship on time, patterns drive engineering investment, delivery rate gets tracked per team. Items that do not ship are theatre, not work; tracked delivery rate is the difference between "we noted it" and "we fixed it."

• Action items ship on time. On-time delivery rate per quarter. Theatre items get caught.
• Patterns drive engineering investment. Engineering project per pattern. Recurring contributing factors become architecture work.
• Tracked delivery rate. Percent-shipped-on-time metric per team. Drives "can we trust this team's commitments?" answer.
• Named owner per pattern. Responsible engineer per pattern. Catches "noted but never acted on" entries.

Compound

Compounding is the visible payoff of the loop. Year over year, MTTR declines and incident frequency drops as systemic fixes land. If neither happens, the loop is broken; capture-without-apply is the most common failure mode and produces a beautiful archive of postmortems that nobody fixed.

• MTTR declines year over year. Trajectory per year. Visible payoff of the loop.
• Incident frequency drops. Count trajectory per year. Systemic fixes shift the baseline downward.
• Diagnostic check if neither happens. Loop broken signal. Investigate the gap; capture-without-apply is the common failure mode.
• Published metrics per year. Visible MTTR and frequency report. Supports honest organisational learning.