LEARN. PRACTICE. GROW.Choose uncertainty that changes a decision
In the fictional Mosaic case, an exception tool has already been purchased. States can still be configured, support prepared, and shift handover improved. The sponsor requests a report confirming the investment. The Product Owner should clarify which decisions remain open and what information could change them. With two days available, uncertainty about pending-work transfer deserves a concrete question: what must the next operator understand to resume work? A supplier demonstration may explain features but does not itself show the affected shift’s experience. Choose activities answering the question, plan access to relevant people, and agree how findings will enter the decision. GOV.UK guidance is a methodological reference; its suggested cadences and participant counts are not universal Scrum rules or bank procedures. This course establishes no sample size guaranteeing representativeness.
Observe without adding a conclusion
In the Lume case, five people use a prototype to distinguish a correctable rejection from an external dependency. Three complete unaided; two find the menu only after guidance. Record the intervention and where they became stuck. The result is not five independent completions. A note such as “opened the detail twice” describes behavior; “distrusts the system” adds an interpretation needing clarification. Group observations with context, discuss interpretations with the team, and retain signals contradicting the preferred solution. Four people may like a label while three understand it in an operationally incorrect way. Preference does not settle comprehension. Ten notes also do not necessarily represent ten people: eight may be attempts by the same person. That repetition may indicate persistent difficulty but should remain linked to research identity and context without inventing independent participants or exposing unnecessary personal data.
Prepare participation and recognize limits
If the problem occurs at night and research includes only the day shift, more daytime interviews do not automatically cover night routing. Prepare participation from the relevant context. A screen-reader user needs materials allowing observation of the intended interaction; a moderator reading images does not demonstrate independent navigation. A prototype without a backend can inform state comprehension but does not demonstrate security, load behavior, or operational readiness. Use entirely fictional data for our rehearsal: there is no implied authorization to move customer data. State the contextual limitation of those data. Two difficulties among five convenience-recruited people describe that sample; they do not establish that 40% of the entire population will fail. Keep impact, evidence, and uncertainty together. If night-shift access is missing, work on the observed improvement while deferring the generalization depending on that access, rather than manufacturing certainty or abandoning all learning.
Lume analysis workshop
Prepare 40 minutes with a Product Owner, researcher, APS, development, and observer. Use ten minutes to distinguish the open decision, question, and context. During the next ten, introduce fictional notes: three independent completions, two assisted completions, and eight repeated attempts by the same person. Request a synthesis that does not mix units. Use another ten minutes to discuss label interpretation and formulate a verifiable improvement option. Close with ten minutes to plan the next round and debrief. The deliverable is a sheet with observation, interpretation, uncertainty, improvement option, and expected evidence. The observer marks “not observed,” “observed with help,” or “observed in rehearsal without help” for each behavior, without an aggregate grade or certification. This workshop rehearses synthetic-material analysis; it is not research with actual users. The guide has been written, but human-participant execution and independent specialist review remain outstanding.
LUME GUIDE | 40 min: 10 + 10 + 10 + 10
Decision still open:
Question / context / relevant people:
Literal observation / source:
Distinct person or repetition:
Task / independent or assisted completion / assistance given:
Interpretation / alternative explanations:
Prototype and sample limits:
Improvement option / expected evidence:
Next round / owner / date:
Observation: not observed | with help | in rehearsal without help“Found the menu after guidance” supports assisted completion. The observation should not be rewritten as independent success to improve the report.
Common pitfalls
Assistance as independence; notes as people; preference as comprehension; prototype as production; research as mandatory confirmation.
Related topics: User research · Synthesis and backlog · Observable criteria
Useful discovery retains observation, exposes uncertainty, and informs a decision that can still change.
Reference: Plan user research for your service · Scrum Guide November2020; EBM May2024; primary product practice reviewed 2026-09-30