LEARN. PRACTICE. GROW.Concept and mechanism
A hypothesis connects a change with an expected effect. “If the alert arrives before the operator starts searching, clarification calls will decrease” can be explored with defined context and measures. Choose the method according to uncertainty: interviews help understand experiences and needs; prototype tasks may reveal label interpretation; a usable pilot enables behavior observation in a situation closer to use. Compare cost, risk, and learning quality. An interview does not demonstrate production performance, and a prototype need not serve the same operational purpose as an increment. Do not automatically choose the most expensive or cheapest experiment without knowing what must be learned.
Guided application
Before the experiment, define the hypothesis, group, expected behavior, and analysis criteria. During a test, avoid indicating the correct path before observing the participant. If help is necessary to continue, record it: assisted completion does not establish independent navigation. If four of six operators confuse labels, there is difficulty to investigate, but not a representative percentage for the entire population. A result that does not confirm benefit can guide the next hypothesis or prevent expensive expansion. When an alert arrives late, evaluate delivery timing before concluding no alert could have value. Retain observations and limitations so the decision can be revisited.
Everyone finishing after receiving instructions does not establish that everyone understood the flow unaided.
Common pitfalls
Prototype as production; success defined after data; unrecorded assistance; small sample as certainty.
Related topics: Backlog, refinement, and outcomes · Costs, value, and segment measures
An experiment should produce enough learning for the next choice with explicit limitations.
Reference: Using moderated usability testing · No official exam required; CSPO learning objectives January 2022 (formatted February 2024); Scrum Guide November 2020