define-ux-success-metrics

Start with the decision each metric feeds, then pick a framework (HEART for consumer products, CASTLE for mandated workplace tools). Choose 2–4 categories, derive goal → signal → metric for each, assign an instrument, and set baselines and targets before build starts.

SUS is a 10-item standardized questionnaire measuring perceived usability of a whole product or flow, scored 0–100. SEQ is a single 7-point question asked immediately after one task to measure its perceived difficulty. Both must be administered verbatim.

NPS alone is too coarse for design decisions: its binning destroys information, it is gameable, and it cannot tell you what to fix. If required, pair every NPS number with task success rate and a perceived-usability score like SUS or SEQ.

No. With n=5 the margin of error swamps the measurement, so '3 of 5 succeeded' is an insight prompt, not a 60% success rate. Quantitative UX metrics need roughly 40 participants, and trendable benchmark waves need about 35 or more.

Choose 2–4 metrics tied to critical tasks, establish a baseline wave with about 35+ participants, freeze the protocol (task wording, instruments, screener, metric definitions), and re-measure on a fixed cadence. If the product change breaks a task, version it and start a new baseline.

Do not use it for analyzing usability findings (analyze-usability-data), designing A/B tests or judging statistical significance (use-quantitative-evidence), measuring AI output quality (write-ai-evals), or writing custom survey questions (write-survey).