Metrics driven development
summary
Shows how JustEat ties product decisions to concrete signals: push business counters (logins, orders), measure timers and gauges for APIs and boxes, ship logs to ELK/Logstash instead of RDP, and hook StatsD/Graphite/Grafana into alerts. Uses feature toggles for graduated rollouts and pushes fake load (50% of projected traffic) so a predicted 1,500 orders/min is tested today. Triggers simple production alerts (success/attempt ratio < 0.7–0.8) and uses graphs to choose what to build next.
AI-written summary · about
chapters (17)
- 0:00Introduction and speaker background
- 1:05Purpose of the talk
- 2:29Frustration with unclear feature requests
- 3:52Domain-driven design challenges
- 5:15What metrics-driven development means
- 6:39Benefits of metrics in production
- 9:05Types of metrics: business, application, infrastructure
- 11:56Changing company culture with metrics
- 13:19Setting up metrics collection tools
- 15:07Iterating and failing fast with metrics
- 16:35Feature toggling and canary releases
- 17:19Alerting based on user behavior
- 18:45Single source of truth with logs
- 20:11Data-driven decision making
- 21:54Fake load testing for scalability
- 24:01Demo of metrics graphs
- 25:54Q&A session
related talks
36mMetrics - What's your code actually doing?
24mMetrics-driven development
37mYou Can't Optimize What You Can't Measure
12mUnlocking the Secrets of Client Data
21mData-driven approaches in a technology startup
33mAnalytics Data Driven Testing
28mProduct metrics for developers
24mHow to monitor your business, and not just your pods