Justin Rohrman on CAST 2014
summary
Argues that modern testing metrics are rooted in social‑science measurement mistakes and that common indicators — defect counts, test coverage, code churn — often explain less than 10% of post‑release failure variance. Demonstrates with historical examples and replicated analyses, shows tools used (surveys, log analysis, basic statistical models) and offers practical remedies such as structured qualitative observation, designed experiments and lightweight Bayesian methods that were peer reviewed at CAST.
AI-written summary · about
related talks
34mesting in modern times: A story about quality and value
43mRegression Obsession
06mLightning Talk: Automated Testing Done Right
39mThe Good Testing, The Bad Testing and The Ugly Testing
45mComplexity Thinking, Cynefin & Why Testing Metrics All Suck
01h 13mPNSQC 2013 - Lee Copeland
28mYour Testing Strategy is broken - lets fix it!
21mWhen a platform isn’t enough: Building an experimentation culture