Apdex (Application Performance Index) collapses a whole latency distribution to a single number between 0 and 1. Pick a target time T that means "fast enough" — requests at or under T are satisfied, requests between T and 4T are tolerating, and everything above 4T is frustrated. The score is (satisfied + tolerating/2) / total. Enter counts directly or paste a list of latencies and the tool bins them for you.
Zone breakdown
Reference bands
| Zone | Range | Weight |
|---|---|---|
| Satisfied | – | 1.0 |
| Tolerating | – | 0.5 |
| Frustrated | – | 0.0 |
Apdex was standardised in 2004 by an industry group (New Relic, HP, Compuware and others) as a way to summarise the user-perceived responsiveness of a web application in one number that is safe to average across services and teams. The three-zone bucketing is deliberately coarse — it maps to the classic fast enough, slow but usable, abandoned mental model — and the 4× tolerating multiplier reflects Jakob Nielsen's usability finding that response times above roughly four seconds are what break users' flow. Modern SRE tooling still exposes it: Datadog, New Relic, Dynatrace, and Grafana all offer Apdex-style widgets, and it remains a common SLO target for interactive user-facing services (typically ≥ 0.90). The score's weakness is the same as its strength: any bucketing scheme hides tail-latency drift and multi-modal distributions, so pair it with a p95 / p99 line chart.