Improving Auto Pause for Everyone
Article Summary
Strava's engineering team faced a deceptively hard problem: how do you automatically detect when a runner stops moving without draining their battery or showing the wrong pace?
The Strava team rebuilt their auto-pause feature from scratch to work across all devices, not just the iPhone 5s. They had to solve fundamentally different problems for cyclists (who want reliability) versus runners (who need sub-second accuracy).
Key Takeaways
- GPS-based auto-pause took 5+ seconds to detect stops, making pace calculations off by 15-25 seconds per mile
- Switched to accelerometer data measuring 'jerk' (acceleration derivative) to detect running motion within 1 second
- Cyclists get inclusive 10-second pause threshold; runners get responsive motion detection that ignores walking
- Final accuracy: within 5-10 seconds per mile even with 2-3 stops, down from 15-25 seconds with GPS alone
Critical Insight
By combining accelerometer motion detection with GPS fallbacks, Strava achieved sub-second pause/resume accuracy while maintaining the same stats from recording to upload.
