What Is Apdex Score? Measuring User Satisfaction

Your checkout page responds in 200 ms for most shoppers, but every few minutes someone waits eight seconds. The average looks fine. Apdex exists because averages hide that kind of pain. The Application Performance Index (Apdex) is an open standard that converts response time measurements into a single score between 0 and 1, where 1 means every user was satisfied and 0 means none were.

The formula works from a single threshold you choose, called T. Any request faster than T counts as satisfied. Requests between T and 4T count as tolerating (weighted at 0.5). Anything slower than 4T, or any server error, counts as frustrated (weighted at 0). The math: Apdex = (Satisfied + Tolerating / 2) / Total. With T set to 500 ms and 200 requests where 170 were under 500 ms, 20 fell between 500 ms and 2 seconds, and 10 exceeded 2 seconds, the score comes out to 0.90.

What the score actually tells you

An Apdex of 0.94 or above is excellent. Between 0.85 and 0.93 is good, which is where most production applications sit. Below 0.70 means a significant share of users are waiting too long. The score is useful as an SLO target ("maintain Apdex above 0.90 for the API") because it captures user-facing impact in a way that raw P99 latency numbers do not. But it has blind spots. Apdex ignores error rates unless the tool explicitly maps errors to the frustrated bucket. It also hides the shape of the distribution: two services scoring 0.85 can have completely different latency profiles. And setting T too high inflates the score without improving the actual experience.

Where Apdex fits alongside uptime monitoring

Apdex is typically calculated by APM tools (New Relic, Datadog, Dynatrace) that instrument application code and measure per-transaction response times from inside the app. Uptime monitoring operates from the outside: it checks endpoints at regular intervals and records whether they respond and how fast. The two are complementary. Uptime monitoring catches outages and external slowdowns that Apdex alone would miss (a site that is completely down has no requests to score). Apdex catches internal performance degradation that external checks at 1-minute intervals would not detect. Teams that care about both availability and performance typically run both.

Related terms: SLO, SLI, P99 latency, uptime, SLA