About APDEX Scorer

Calculate Application Performance Index (APDEX) scores with satisfaction thresholds. Part of the DevTools Surf developer suite. Browse more tools in the Developer Utilities collection.

Use Cases

• Track user satisfaction across API endpoints after a performance regression
• Set SLO targets in terms of Apdex score rather than raw latency percentiles
• Compare performance of microservices using a single normalized metric
• Report performance to non-technical stakeholders using a 0-1 score

Tips

• Set T (the satisfied threshold) to your 95th percentile response time target, not the average — Apdex rewards fast performance, not just acceptable
• Input raw response time samples; the tool bins them into satisfied (<T), tolerating (T-4T), and frustrated (>4T) buckets automatically
• Compare Apdex scores before and after deployments to quantify user experience impact objectively

Fun Facts

• APDEX (Application Performance Index) was defined by an industry consortium in 2007. The formula is (Satisfied + 0.5 * Tolerating) / Total — producing a score from 0 to 1.
• An Apdex score of 0.94 is considered 'excellent'. Scores below 0.85 indicate users are noticeably experiencing degraded performance and churn risk increases.
• Apdex was designed partly because raw latency averages are misleading: a system where 95% of requests take 100ms and 5% take 10 seconds has a good average but terrible user experience.

FAQ

What T value should I use?

T is application-specific. A common starting point is 500ms for web pages and 100ms for API calls. Set T based on user research or your SLA targets, not industry averages.

Is Apdex better than p95 latency?

They answer different questions. p95 tells you the worst-case experience for the 95th percentile user. Apdex summarizes the distribution of experiences across all users into a single satisfaction score. Use both.

What Apdex score is acceptable?

0.94+ is excellent; 0.85-0.93 is good; 0.70-0.84 is fair (users notice); below 0.70 is poor — at this level, performance is likely causing measurable abandonment.