Transmissions From the Tweet Factory: How Observability Transformed Mobile Performance at Twitter
Talkat Observability Day 2025 · on Tuesday, 2025-04-01· at 14:40· Level 1 | Hall Entrance S10 | Room B·
summary
Twitter builds on PerfTown logs then develops Production Client Tracing (PCT) optimized for mobile (spans, durable storage, minimal overhead) and integrates with Kafka, Druid and existing analytics; rollout takes years (2015 → PCT 2019 → established by mid‑2022). Observability yields concrete wins: startup and network optimizations improve monetizable DAUs, a surprising cohort of ~50,000 daily users traced to a preinstalled Japanese app version, and a new golden metric — user‑operation failure rate — directly links client perf to product KPIs.
AI-written summary · about
chapters (13)
- 0:00Mobile observability at Twitter
- 1:08Mobile performance challenges 2015
- 3:20Perf Town metrics basics
- 5:31Metrics limitations and need
- 7:00Why tools failed for mobile
- 9:08Production Client Tracing design
- 11:39Rolling out mobile tracing challenges
- 13:27Mobile observability by 2022
- 14:55User behavior patterns case study
- 16:43User-centric failure rate metric
- 19:34From app to user experience
- 20:59Key takeaways Twitter observability
- 21:40Future of mobile observability
related talks
30miOS Observability & Performance Metrics
26mObservability of your app in production
34mClient Tracing At Scale
19mOops my app performance sucks... what if observability could be a game changer?
28mScaling Observability for Real-Time Personalization Engines
28mMobile Application Performance Done The Right Way
19mUser-Focused Observability 101 For Android Devs
20mObservability and Open Telemetry