You Get What You Measure—Why Metrics Are Important
summary
Shows concrete failures and fixes: unit mismatches cost a Mars probe and pounds vs kilograms causes an Air Canada dual-engine flameout; availability expressed as four nines = 52 minutes/yr makes teams complacent so downtime framing replaces nines; search pages render only ads while UX 200s hide loss, prompting a no-algo no-results gauge and session-success counters; blind spots at high percentiles and CDN blackholing drive added anomaly detection and full-fidelity requests for more accurate alerting and post-mortems.
AI-written summary · about
chapters (13)
- 0:00Metrics introduction
- 0:49NASA Mars probe unit mismatch
- 2:14Air Canada fuel gauge error
- 5:47Availability metrics explained
- 8:35Pitfalls in metric calculations
- 9:16Danger of single metrics
- 11:43Balancing product metrics
- 13:30Missing real user impact
- 15:16High percentile blind spots
- 16:44Page load time gaming fixes
- 18:48Metric gaming examples
- 21:36Closing on metric quality
- 22:01Q&A
related talks
01h 01mFinding Metrics that matter and using them safely
25mWhen Metrics Lie
36mMetrics - What's your code actually doing?
37mYou Can't Optimize What You Can't Measure
27mMetrics that Matter - Moving from Easy to Impactful
10mOf Mice and Metrics
56m"Mirror mirror on the wall, what is the vainest metric of them all?
01h 01mKeynote - Starting a metrics program