Error rate is the share of items or interactions that are wrong, calculated as errors divided by the total checked. It's the flip side of accuracy, 95% accuracy is a 5% error rate.

‍

‍Why it matters

The two numbers describe the same thing, but error rate tends to start a different conversation. Saying 5% error makes people ask what's behind that 5%, while 95% accurate can sound finished. It also scales in a way that's easy to lose sight of, 5% of a million items is 50,000 mistakes. In support specifically, error rate doesn't always move with satisfaction, a friendly, well-liked agent can still confidently give wrong information, which is why the two get tracked separately rather than assumed to move together.

‍

‍How to track it usefully

Error rate is most useful broken down by category, type, or person rather than reported as one blended number, since a 5% overall rate can hide one category failing far more often, sitting inside a large volume of easy items that are almost always right. In support, teams typically split it further into wrong information versus wrong action, since each points to a different fix, a knowledge base gap versus a process issue.

‍

‍When a low error rate can mislead

Error rate can look great on unbalanced data. If only 1% of items are actually violations or mistakes, a system or person that flags nothing is wrong just 1% of the time, while catching zero real problems. This is often called the accuracy paradox, and it's a big reason metrics like precision and recall exist as a separate check alongside a blended error rate.