Turn raw pipeline outcomes into an interpretable error profile. Enter evaluated and erroneous outcomes to calculate the error rate, complementary success rate, errors per 1,000, and an approximate 95% Wilson interval. Use it to compare batches while keeping both rate and sample size visible.
Success rate is 100% minus the error rate. The uncertainty range uses a 95% Wilson score interval.
A lower rate means fewer observed failures, but severity and type still matter. Track the numerator, denominator, and error categories alongside the percentage.
Do not compare two rates without considering sample size, sampling method, and whether the error definitions are identical.
For 37 errors in 1,000 evaluated outcomes, the error rate is 3.70%, the success rate is 96.30%, and there are 37 errors per 1,000.
Divide failed or inaccurate outcomes by all evaluated outcomes, then multiply by 100.
An error rate is undefined with a zero denominator, so the calculator asks for at least one evaluated record.
Yes when every evaluated outcome is classified only as correct or incorrect. More complex scoring systems may not follow that identity.
The normalized count makes rates easier to compare across batches with different volumes.
No. Sample representativeness, error severity, drift, and the confidence interval around the rate also matter.
| Output | Use |
|---|---|
| Error rate | Share of evaluated outcomes that failed |
| Per 1,000 | Normalized comparison across batch sizes |
| 95% interval | Sampling uncertainty around the observed rate |