Debugging Android ANRs at Scale: Lessons from Building a Mobile Profiler

Every Android developer knows the frustration: Your launch times are trending up, your ANR rate is spiking, and your Play Store rating is at risk. Your observability tools tell you what broke but not why.
Android’s ProfilingManager API opens the door to collecting production profiling data like stack samples, traces, and heap dumps directly on user devices. But raw Perfetto traces alone aren’t enough. The real challenge is turning that data into actionable insights.
In this talk, we’ll share how Datadog built a production profiling pipeline on top of the ProfilingManager API in collaboration with Google’s Android team. From collecting traces at scale to linking profiling data with real user sessions, surfacing flame graphs and call stacks, and pinpointing the methods behind slow launches or ANR spikes. We’ll walk through real debugging examples, preview what’s next for Android profiling, and show how these techniques can fit into your own debugging workflow.