Creating Airbnb's Page Performance Score
Article Summary
Airbnb ditched single-metric performance tracking and built something better. Here's how they unified web, iOS, and Android performance into one 0-100 score.
Andrew Scheuermann from Airbnb's engineering team explains how they evolved from tracking one metric (Time To Airbnb Interactive) to a comprehensive Page Performance Score that works across all platforms. The system combines multiple user-centric metrics into a single, comparable number.
Key Takeaways
- PPS combines 5-7 platform-specific metrics into one 0-100 score
- Metrics split between initial load (TTFCP, TTFMP) and post-load (FID, TBT, CLS)
- Formula uses weighted curves: TTFCP gets 35%, FID 30%, TTFMP 15%
- Weighted Average Score prioritizes high-traffic pages across hundreds of pages
- System designed to evolve: added Cumulative Layout Shift in 2019 seamlessly
Airbnb created a flexible, multi-metric performance scoring system that lets them compare progress across platforms while maintaining the simplicity of tracking a single number.
About This Article
Airbnb relied on a single TTAI metric that didn't work well across different platforms. Each platform had its own baseline, and the definition of 'interactive' kept changing, making page comparisons unreliable. Sometimes engagement metrics would improve while TTAI stayed flat.
Andrew Scheuermann's team created PPS with logarithmic scoring curves and platform-specific thresholds. They set Good above 0.7, Poor below 0.5, and Moderate in between. They then weighted the metrics by analyzing A/B test data to match what the company's internal engagement metrics showed.
Airbnb moved from tracking time in seconds to a unified 0-100 score system. The system adapted over time, and when they added Cumulative Layout Shift in 2019, it didn't break how the organization aligned on metrics or reported results.
