Patterns, Not Categories: Learning Across Incidents
summary
Argues that category counts and statistical bucketing often fail to produce actionable fixes, citing a 2017 Microsoft cross-division analysis that collapses under inconsistent postmortem data and a Spotify “incident archeology” example that still finds ambiguity. Proposes Christopher Alexander–style pattern languages instead, with concrete patterns like clumsy automation and multi-party problem, tools shown include Grafana dashboards, a Federated Wiki pattern page and flashcard packs, and a handout plus live survey already collecting patterns with a public repository planned.
AI-written summary · about
chapters (15)
- 0:00Patterns, Not Categories: Learning Across Incidents
- 5:29Do the categories serve us?
- 9:16Repeat Incidents
- 11:41Jumping into An Incident
- 14:23Pattern Concept Abstraction
- 19:41You Are (Part of) The System
- 21:50Federated Decisions
- 23:28Pair Programming Spread expertise and reduce dependency
- 27:43Working at Cross Purposes Law
- 27:56Fragmentation Across Unit Boundaries
- 28:54Multi-Party Problem
- 31:12Clumsy Automation Pattern
- 32:51Workarounds
- 35:00Pattern Finding
- 40:34Predictive Power
related talks
07mOrganizational Design Forum Report Out
55mSpaker Panel - Day 2
29mBreak Free of the Template: Incident Writeups They Want to Read
47m8 Cloud Design Patterns you ought to know.
36mFrom Anti-pattern to Pattern Driven Architecture Definition
36mApplication Architecture Patterns
25mContinuous Delivery Design Patterns
34mTaking Human Performance Seriously in Software