Learning From Near Misses¶

Why It Matters¶

Organizations wait for disasters to investigate seriously. But near misses contain nearly all the causal information of actual accidents—the only difference is luck in the final outcome. The ratio of near misses to accidents can be hundreds to one (Heinrich's triangle suggests 300:29:1 for near misses:minor injuries:major injuries). Treating near misses seriously multiplies learning opportunities and enables prevention rather than reaction.

The Key Move¶

Establish that near misses deserve investigation comparable to actual accidents—not dismissal because "nothing happened." Create reporting systems that encourage near miss reporting without punishment. For each near miss, perform the same analysis as accident investigation: identify necessary conditions, trace contributing factors, check for common causes across multiple near misses, distinguish individual versus systemic factors. Look for patterns: multiple near misses of the same type indicate systemic vulnerability. Ask: "What would have to change for this to become an actual accident?" Then prevent that change.

Classic Application¶

Aviation's ASRS (Aviation Safety Reporting System). Pilots submit near miss reports (loss of separation, unstabilized approaches, equipment anomalies) without punishment. NASA analyzes patterns, issues safety alerts, and feeds back learning to the aviation community. Estimated to have prevented hundreds of accidents by learning from thousands of near misses annually. The system works because reporting is non-punitive and actually produces visible safety improvements.

Surprising Application¶

Software deployment near misses. Production deployment almost caused an outage but was caught by monitoring before user impact. Most organizations dismiss this ("crisis averted, moving on"). Instead: investigate why deployment almost failed, what conditions made it nearly catastrophic, whether other deployments have similar vulnerabilities, and what systematic changes would prevent recurrence. Google's postmortem culture explicitly includes "near misses" for this reason.

Failure Modes¶

Requires overcoming hindsight bias—people dismiss near misses because "nothing bad happened." Reporting systems fail if punishment culture makes reporting risky. Can create excessive investigation burden if every anomaly gets full accident-level analysis (requires triage). Near miss definition is fuzzy—how close to accident counts as near miss? Organizations may game metrics by calling everything a "near miss" to appear safety-conscious. Patterns may not emerge if reports aren't systematically analyzed.

Go Deeper¶

Reason, Managing the Risks of Organizational Accidents, Chapter 12; ASRS Database & Reporting System (https://asrs.arc.nasa.gov); Tinsley, Dillon & Cronin, "How Near-Miss Events Amplify or Attenuate Risky Decision Making," Management Science 58, 2012.