How Uber Standardized Mobile Analytics for Cross-Platform Insights

Ben Hjerrild and team at Uber reveal how they cut impression inflation by 30% and slashed analytics code by half. Their secret? Moving instrumentation logic out of feature code and into the platform layer.

Uber's Mobile Data Platform team tackled a messy analytics problem: 40% of events were ad-hoc custom logs, metadata was duplicated everywhere, and iOS/Android emitted events differently. They redesigned their entire mobile analytics stack to standardize event definitions, automate metadata collection, and ensure cross-platform consistency across all Uber apps.

Key Takeaways

• Reduced transient view impressions by 30% with stricter visibility rules
• Cut instrumentation code from dozens of lines to just a few per event
• Standardized three core events: tap, impression (50% visible for 500ms), and scroll
• Sample 0.1% of sessions to maintain visibility into disabled or unmapped events
• AnalyticsBuilder classes encapsulate emission logic, separating concerns from UI code

Their next move with componentization promises to cut event mapping effort in half again using a single UUID per component instead of per event.