Unit 2 · Lesson 6 of 20
Choose evidence for the question
Match an evidence source to the uncertainty it can reduce.
01 / Learn
The idea
Different evidence answers different questions. Analytics can show where and how often a measured event occurs, but usually not why. A task observation can reveal how someone interprets a flow, but a small qualitative study does not estimate a population rate. Interviews can uncover expectations and context, while support tickets can expose recurring complaints with selection bias.
Define the decision before selecting a method: what might change if the answer is A versus B? Then collect enough evidence to distinguish plausible causes. For an account-recovery exit, inspect the event definition and technical failures, observe representative people attempting recovery, and compare support themes. Triangulating sources strengthens a conclusion when they agree; conflicting evidence is a reason to investigate, not to pick a favorite number.
A single metric or a few quotes provide a full explanation. Each source has blind spots that should be named in the recommendation.
02 / See the reasoning
Worked example
Fictional scenario
A product team sees a high exit rate on an account-recovery page and assumes people dislike its layout.
A team finds that 'exit' includes people who successfully opened their email app to retrieve a code. The metric is not a clean abandonment measure. An expert first repairs the event definition, then tests whether users can return to the flow after switching apps. The change in measurement alters the design question.
Try a decision →03 / Decide
Guided decision
The team must decide whether to redesign the recovery page next sprint. Which evidence plan best supports that decision?
Feedback on your choice
What could change the answer: If a verified event already distinguishes email-switch success from abandonment, an experiment or narrower usability study becomes more appropriate.
This feedback helps you examine the decision; it does not grade your written reasoning.
Apply it somewhere new →04 / Apply elsewhere
Independent application
Fictional scenario
A voice assistant's help command is used rarely. The team assumes people do not need help.
Write a research plan that separates discoverability, need, recognition failure, and alternative help-seeking. Explain what logs, observation, and interviews can each establish.
Make: A decision-question-to-evidence matrix with limitations.
Each method is tied to a question, measurement validity is checked, and the learner avoids turning absence of use into absence of need.
Lesson complete
You applied the method in a new context.
You studied the idea, tested a decision, and made an artifact for a new context. Keep the reasoning for your next UX question.
05 / Keep working