Metrics & Measurement
What Is Error Rate in Load Testing?
Error rate is the share of requests in a test that failed — connection errors, timeouts, unexpected status codes, or responses that failed an assertion.
Also known as: Failure Rate, Error Percentage
Check it before anything else
Error rate is the first number to read in any result, because it determines whether the rest of the numbers mean anything. A run reporting a 40 ms P95 and a 60% error rate is not a fast system; it is a system rejecting most of its traffic quickly.
Errors are usually cheaper to serve than successes — a 500 from a failed database call returns long before a successful query would. So rising errors actively improve your response-time percentiles, which is exactly why reading them without the error rate is dangerous.
What counts as an error
The definition is yours to set, and setting it too narrowly is the usual mistake.
- Transport failures — connection refused, reset, DNS failure, TLS handshake failure.
- Timeouts, which are worth counting separately since they indicate saturation rather than a bug.
- Unexpected status codes: 5xx always, and 4xx when the scenario should not be producing them.
- Responses that passed the status check but failed an assertion — an HTTP 200 rendering an error page is still a failure.
The curve matters more than the number
A single figure for the whole run averages away the thing you need. Plot error rate against time and against load, and the shape tells you which failure you are looking at.
Errors clustered at the start usually mean warm-up or connection-pool ramp. Errors that begin at a specific load level mark the saturation point. Errors that climb steadily through a flat load are a leak or an exhaustion problem, and belong to a soak test investigation.