on Picking the Right Analytics Metrics
summary
Argues that a good performance metric meets four concrete criteria and offers how to apply them: simple, reproducible, credible, actionable is the proposed checklist, with the NFL quarterback rating used as an example of a noncredible metric because its components are arbitrary. Recommends separate technical and business metric sets rather than an omnibus measure, advises periodic reevaluation and to retire it when not actionable, and emphasizes reproducibility from input data.
AI-written summary · about
chapters (5)
- 0:00Start
- 0:07Metric Should Be Simple Reproducible Credible and Actionable
- 1:20Should a Metric Be Able To Address both Technical and Business Aspects
- 2:32Do Metrics Have a Life Cycle
- 2:49How Do You Know When It's Time To Let It Go
related talks
38mThe Metric System: Making Correct Performance Measurements
35mHonest Performance Testing
47mMobile Performance: Decoding Metrics
12mPerformance In Context
24mHow To Manage Performance Like a Product
38mPerformance 101
38mCreating & Scaling Meaningful Performance Tests in CI
31mPrecision Over Proliferation: SRE Approach for Leaner, Smarter and Data-Driven..