What it is
A method where representative users attempt realistic tasks with a product or prototype while an observer watches what they do (and, when thinking aloud, hears why). It surfaces behavioral evidence of where the design works and where it breaks down.
When to use it
Whenever you have something a person can attempt, a task flow, a prototype, or a live product, and you need evidence that people can actually complete key tasks. Fits Prototype (evaluate a direction before building) and Implement (validate the built experience); the planning and task design draw on Immerse understanding.
Why (value)
Reduces the risk of shipping a flow people cannot complete. Turns opinions and internal debate into observed behavior, and produces specific, prioritizable findings tied to real tasks.
How to run it
- Define the tasks and success measures before sessions begin (distinguish observed behavior from self-report).
- Recruit participants who represent the relevant audience.
- Have each participant attempt the tasks; observe behavior, ask them to think aloud, avoid leading them.
- Capture where they hesitate, err, or fail, and note successes as well as problems.
- Synthesize across sessions into findings and severity, then recommendations.
Inputs and outputs
- Inputs: a testable artifact (prototype or live product), representative tasks, success criteria, recruited participants.
- Outputs: observed task successes/failures, prioritized usability findings, recommendations, supporting clips or notes.
Time and effort
Medium. Roughly 5 to 8 participants per round for qualitative discovery of major issues. Can run moderated (remote or in person) or unmoderated with a tool. Plan for recruiting lead time.
Evidence and limitations
Yields direct-user behavioral evidence, the strongest common evidence type for task performance. It shows what users do, not necessarily what they prefer or would do long-term. Small qualitative samples identify issues but do not produce statistically reliable rates. It is distinct from expert review, which is not user research. See Evidence and limitations.
Tips and pitfalls
- Do not lead or rescue the participant; let them struggle enough to reveal the issue.
- Test tasks, not opinions; avoid asking “do you like this?”
- Define success up front so analysis is not purely retrospective impression.
- Separate what users say from what they actually did.
Variations
Moderated vs. unmoderated; remote vs. in person; think-aloud vs. silent observation with retrospective probing; benchmark (quant metrics) vs. formative (diagnostic) testing.
Related methods
- Heuristic evaluation
- Think-aloud protocol
- First-click testing
Need help planning research or facilitating this method?
Gumshoe helps teams connect research, collaborative workshops and service design to the decisions they need to make.
Sources and further reading
- NN/g, “Which UX Research Methods?” (landscape of methods).
- 18F Methods (US government, public domain).
Explore related methods
- Heuristic evaluation — Have evaluators inspect an interface against usability principles to catch problems fast.
- First-click testing — Measure where people click first on a task, since that first move strongly predicts success.
