A zero-error interface may just have converted errors into abandonment

What it is

Error rates fall for two reasons: the interface got better, or users never reached the place where errors happen—they abandoned before the hard step. In the second case the error data is pristine while completion collapses. "Zero errors" is not usability evidence on its own; it must be read together with abandonment and completion.

Why it happens

An error is a record of "tried and failed," and abandonment deletes the "tried," so the count zeroes out. Defensive use (not daring to press uncertain buttons) and entries too hard to find (never reaching the error point) both manufacture this silent improvement. A stealthier form is "errors converted into detours": users finish by re-entering everything instead of fixing one field—error rate down, burden up. Any monitor watching error rate alone reads that degradation as progress.

Studying it

Reporting rules require error rate to ship with arrival, abandonment, and completion: a funnel showing reached-attempted-errored-completed-abandoned per step, where zero errors with low completion is the danger signature. In logs, the last actions before exit distinguish "finished cleanly" from "quit in frustration"; frustration signals (rapid repeated clicks, back-button storms, abnormal dwell) serve as supporting evidence for abandonment typing.

Where it stops holding

Not every low error rate is fake: genuinely easy tasks or expert populations can earn real zeros. "Abandonment" also needs care—normal interruptions (a phone call, a changed mind) look different in logs from frustrated quits and require pattern separation, not blanket counting. The warning mainly targets treating error rate as a standalone optimization target; in systems that report metrics jointly, the risk is hedged by construction.

Applying it

• Make error-rate dashboards show arrival, abandonment, and completion on the same screen; single-metric alerts never fire alone.
• Add a cross-check to release acceptance: when error rate falls but completion doesn't rise, the reason must be explained (usually path changes or abandonment shift).
• Run periodic "hard steps nobody visits" audits: a step never reached may be an entry defect, not absent demand.

Related

• Same group: B5.10.1 Error rate correlates with effectiveness but must be reported separately; their remedies differ · B5.10.2 Errors must be counted separately as recoverable and unrecoverable; merged counts hide severe problems · B5.10.4 Error rate responds to interface changes more sensitively than satisfaction, making it a good regression monitor
• Nearby: H1 Interaction Patterns and Flows · Q2 Usability Testing
• Search terms: silent failure · abandonment rate · funnel analysis