Fairytale vs. Narrative: Telling stories and making decisions with data
summary
Explains concrete failures and fixes: why Facebook flagging abused proves a single metric can be misleading, how using story points mask bugs cost users real outcomes, and why dividing yearly pageviews by applicants fails—so the real fix is to interview applicants for causation. Shows an example forensic workflow (IP timing, rotating addresses) that narrows but does not prove attackers, and recommends hiring statisticians, designing A/B experiments with controls, and collecting events with intent rather than retrofitting numbers.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 2:29Data-driven decisions fail
- 7:49Data vs story in decisions
- 12:23Inventing explanations
- 16:32Single data point dangers
- 23:20Using available data
- 28:59Ignoring context
- 36:24Discounting alternatives
- 41:46Better data narratives
- 46:19Q&A
