67: Measuring Developer Productivity with Diff Authoring Time

This Meta Tech Podcast episode interviews Sarita and Moritz about "Diff Authoring Time" (DAT), a metric they developed to quantify internal developer productivity. The discussion covers DAT's implementation, including its use of operating system and IDE telemetry to track developer activity, and its validation process, which involved multiple iterations and rigorous checks to ensure accuracy and coverage (currently at 87% of eligible diffs). The metric is used for aggregate analysis, such as comparing developer cohorts and A/B testing new features, and is not used for individual performance evaluations. The average DAT is approximately 50 minutes per diff, providing a benchmark for improvement efforts. Future iterations aim to simplify the metric and provide actionable insights for developers.