Dealing with Single Incidents Exceptions

00:24:17 — Anti-Pattern: Single Incidents as Representative Reality

Cottrell identifies a key anti-pattern in engineering management: treating a given incident as representative of some larger reality. Humans are much better at remembering individual incidents, especially extraordinary ones, than at seeing patterns across many events. A single failure can easily stick out above all your successes. This cognitive bias makes the systematic trend data from incident review programs so valuable — without it, evaluation of direct reports and even self-assessment becomes distorted by memorable outliers.

00:25:54 — Flame Graphs for Management: Distribution Over Events

Larson draws an analogy to flame graphs in software debugging: any single slice doesn’t necessarily tell you something useful, but as you take many slices, they start coming together into an accurate picture of what’s actually happening. If you only take a couple of samples, you’ll almost always focus on fixing the wrong problem. Being careful to look at the distribution rather than the singular event is one of the most important things he’s done to improve the quality of his thinking and decision-making.

00:28:25 — Dealing with Exceptions: Core Leadership Skill

Larson argues that dealing with exceptions is a core leadership skill, particularly when designing processes. Good process requires consistency of application, but bad process comes from the needless urge to force everyone into the exact same system while ignoring circumstances. He draws a parallel to software architecture: find the common patterns, then make them extensible to handle what reality brings you. The challenge is making this both effective and fair — not just a system that gives the results people wanted to have, but one that is genuinely just.