Start with the question. Choose the metric. Keep the context.
Can users complete what they came to do?
Observe a clearly defined task and record complete, partial and failed attempts. Establish the same success criteria before comparing designs.
Fewer failed journeys may be associated with fewer support interactions.
Use consistent definitions and comparable samples. A signal alone does not establish causality.
How much effort does the journey require?
Measure the time needed to complete a specific task. Report the median alongside the mean when outliers are present, and distinguish successful from failed attempts.
Faster workflows may indicate operational time savings.
Use consistent definitions and comparable samples. A signal alone does not establish causality.
Where does the experience break down?
Define what counts as an error and its severity. Observe the number of errors relative to consistent opportunities for error.
Fewer errors can help identify opportunities to reduce rework.
Use consistent definitions and comparable samples. A signal alone does not establish causality.
How usable does the experience feel?
Use the standard ten alternating positive and negative statements. Subtract 1 from odd-item scores, subtract even-item scores from 5, sum and multiply by 2.5. SUS is a score, not a percentage.
Perceived usability provides context alongside behavioral evidence.
Use consistent definitions and comparable samples. A signal alone does not establish causality.
Are people using the features that matter?
Define eligible users, a meaningful adoption event and a consistent time window. Distinguish trying a feature from sustained use.
Adoption may be related to product utilization or retention; validate the relationship.
Use consistent definitions and comparable samples. A signal alone does not establish causality.
Are users returning to meaningful interactions?
Define meaningful activity in the context of user goals. Use a fixed period and avoid treating raw clicks or time spent as inherently positive.
Meaningful engagement can provide evidence of sustained product value.
Use consistent definitions and comparable samples. A signal alone does not establish causality.
