Error rate correlates with effectiveness but must be reported separately; their remedies differ

What it is

Error rate correlates with effectiveness (more errors usually mean worse completion quality), but it is an independent dimension: two interfaces with the same completion rate can differ several-fold in error rate, and a falling error rate does not automatically raise effectiveness—errors may just be taking a stealthier form. Measure and report them separately, because fixing them calls for entirely different tools.

Why it happens

Correlation without coincidence exists because errors and "did the task complete" sit on different causal chains: effectiveness is decided by whether the flow goes through; error rate by operation and feedback quality. The remedy split is concrete: raising effectiveness means restructuring the flow (closing breaks, fixing paths); lowering errors means prevention and tolerance design (input constraints, confirmation, undo). Folding error rate into effectiveness misroutes resources—an interface with an intact flow but frequent errors would be misjudged as healthy.

Studying it

Measure errors in their own table: errors per task, proportion of users erring, error-type distribution. Analyze the correlation between error rate and completion quality (accuracy and completeness) to expose "high-error, high-completion" pathological patterns. Attribution research separates error sources: comprehension errors (wrong model), slips (right intention, wrong action), and system errors (product faults)—three remedies, and one blended rate guides none of them.

Where it stops holding

Error identification depends on definition: what counts as an error (invalid clicks? pre-correction intermediate states?) must be agreed in advance, and drifting calibers break time series. Low error rates can reflect trivially easy tasks or overly cautious participants. On infrequent tasks the sample is thin, and stable error rates require pooling across rounds.

Applying it

• Report error rates in their own table with type distribution, never folded into completion narrative.
• Route fixes by error type: comprehension errors to information architecture, slips to prevention mechanisms, system errors to engineering quality.
• Freeze the error caliber for version comparisons; if the caliber changes, re-baseline.

Related

• Same group: B5.10.2 Errors must be counted separately as recoverable and unrecoverable; merged counts hide severe problems · B5.10.3 A zero-error interface may just have converted errors into abandonment · B5.10.4 Error rate responds to interface changes more sensitively than satisfaction, making it a good regression monitor
• Nearby: B5.02 Three Components · Q2 Usability Testing
• Search terms: error rate · error taxonomy · usability metrics