---
name: user-research-and-discovery
description: Plan, conduct, synthesize, or audit generative and qualitative user research, including interviews, contextual inquiry, observation, diary studies, concept evaluation, and usability discovery. Use when a product or service decision needs direct evidence about users' goals, behavior, context, language, barriers, or mental models, and produce a User Research Protocol and Findings. Do not use for market sizing or competitor research, continuous post-use feedback operations, product analytics instrumentation, causal experiment design, or a quantitative prevalence claim.
---

# User Research and Discovery

Produce decision-relevant user evidence without turning a convenient sample,
memorable quotation, or researcher's interpretation into a population claim.
Read [references/research-method.md](references/research-method.md) before
selecting participants, collecting data, or interpreting findings.

## Workflow

1. Bind the study to a decision, target users, context, known evidence,
   assumptions, unknowns, and evidence that could change the decision.
2. Choose the smallest method mix that can observe the relevant behavior or
   elicit the relevant experience. Prefer observation of real work over stated
   preference when the research question concerns behavior.
3. Define a purposive recruitment matrix across material behaviors, access
   needs, experience, context, and risk. State excluded groups and coverage
   limits; do not call a qualitative sample representative.
4. Establish consent, privacy, recording, compensation, safeguarding,
   accessibility, retention, withdrawal, and sensitive-topic handling before
   recruitment. Collect only data required by the study.
5. Write and pilot a protocol. Use neutral prompts, realistic tasks, consistent
   probes, and explicit stopping or escalation rules. Do not coach participants
   toward the intended design.
6. Preserve traceability from session evidence to observations, interpretations,
   themes, and recommendations. Separate what a participant did or said from
   what the researcher inferred.
7. Analyze iteratively across participants. Search for negative cases,
   contradictory segments, researcher effects, non-response, proxy-user error,
   and alternative explanations. Do not manufacture saturation from repetition
   in a narrow sample.
8. Connect findings to the original decision. Rank findings by consequence,
   evidence coverage, and confidence; state unresolved questions and the next
   evidence needed.

## Artifact

Produce a **User Research Protocol and Findings**:

- decision, research questions, users, context, scope, and prior evidence;
- method, recruitment matrix, inclusion and exclusion, consent, privacy,
  accessibility, retention, and study materials;
- session ledger with evidence locators and material deviations;
- findings with supporting and disconfirming observations, affected groups,
  coverage, confidence, and limitations;
- design or service implications, non-claims, unresolved risks, and decision
  handoff.

If only the protocol or only the findings are requested, produce that section
without inventing the missing study phase.

## Boundaries

- Use `market-research-synthesis` for category, competitor, pricing, or demand
  evidence from market sources.
- Use `product-feedback-learning-loop` for recurring post-use feedback intake,
  triage, prioritization, and closure.
- Use `product-analytics-instrumentation-review` for event definitions,
  telemetry coverage, and quantitative behavioral measurement.
- Use `product-experiment-review` for randomized, holdout, switchback, or staged
  product experiments; use `causal-inference-analysis` for causal effect claims.
- Do not generalize prevalence from qualitative evidence. Hand a prevalence
  question to an appropriately sampled quantitative study.