This page is part of the ForgeSDLC knowledge base — an AI-assisted, human-directed methodology for taking product work from concept to production. For the core operating model and vocabulary, see Forge SDLC overview and What is ForgeSDLC?.
User research methods and practices
Purpose: Project-agnostic guidance for understanding users through systematic inquiry — choosing methods, planning ethically, synthesizing findings, and turning evidence into design decisions.
Overview
User research grounds product decisions in observed and reported behavior, not assumptions. It spans generative work (discovering needs and mental models) and evaluative work (testing whether designs work). Good research is planned, transparent about limitations, and connected to action — insights that never influence design or strategy waste participant time and team trust.
Research method taxonomy (2×2)
| Attitudinal (what people say they believe, prefer, or would do) | Behavioral (what people actually do) | |
|---|---|---|
| Qualitative | Examples: interviews, diary studies (self-report), focus groups, card-sort explanations | Examples: contextual inquiry, moderated usability tests, think-aloud sessions, ethnographic observation |
| Quantitative | Examples: attitude surveys, NPS-style scales, preference polls | Examples: analytics, A/B tests, task success rates, click-stream analysis, eye-tracking metrics |
Use the matrix to pair methods: qualitative behavioral observation explains why quantitative behavioral data shows a drop-off; attitudinal surveys at scale complement deep interviews.
Qualitative methods
| Method | Description | When to use | Typical sample | Primary outputs | Effort |
|---|---|---|---|---|---|
| User interviews | One-on-one structured or semi-structured conversations | Exploring mental models, goals, vocabulary; early discovery | 5–12 per segment | Themes, quotes, journey hypotheses | Medium |
| Contextual inquiry | Observation + interview in the user’s real environment | Workflow-heavy domains; understanding tools and interruptions | 4–8 | Task models, environmental constraints | High |
| Diary studies | Participants log experiences over days or weeks | Habits, infrequent events, longitudinal sentiment | 8–20 | Patterns over time, triggers | High |
| Card sorting | Users group labels into categories | IA labels, navigation groupings | 15–30 (quant) or 5–8 (detailed) | Category model, disagreements | Low–medium |
| Tree testing | Find items in a text-only hierarchy | Validate IA without visual design | 30–50+ for stats; smaller for pilots | Findability %, wrong paths | Low–medium |
| Focus groups | Facilitated group discussion | Exploring group norms, reactions to concepts — not for usability | 2–4 groups × 6–8 | Themes, vocabulary, concerns | Medium |
Quantitative methods
| Method | Description | When to use | Typical sample | Primary outputs | Effort |
|---|---|---|---|---|---|
| Surveys | Structured questionnaires at scale | Segment sizing, satisfaction, feature prioritization | 100+ for stable proportions; more for subgroups | Distributions, correlations | Low–medium |
| A/B testing | Randomized exposure to variants | Validate specific UI or copy changes with measurable goals | Powered by MDE and baseline rate | Lift, guardrails | Medium–high |
| Analytics | Instrumented product usage | Funnels, retention, feature adoption | Full population or sampled | Trends, segments, anomalies | Ongoing |
| Task completion rates | Success/fail on defined tasks (lab or unmoderated) | Benchmark usability; compare designs | 20–40+ per variant for stable rates | % success, paths | Medium |
| Click-stream analysis | Sequences of clicks or navigation paths | Diagnose confusion, loops, abandonment | Large event volume | Paths, drop-off points | Medium |
| Eye tracking | Gaze position and fixation metrics | Packaged goods, dense UIs, ad/layout research | Small N + overlays | Heatmaps, fixation order | High |
Research process (flowchart)
Research planning
- Research questions: Separate learning goals (“What frustrates users when onboarding?”) from methods (“Run 8 interviews”). One study should answer a small set of aligned questions.
- Recruitment: Write screeners for behavior and context (not just demographics). Decide sampling: convenience vs quota vs random — document bias. Include edge cases when risk is high (compliance, safety, money).
- Ethics: Informed consent (recording, data use, withdrawal). Fair incentives proportional to burden. Extra care for vulnerable populations (minors, health, financial distress) — involve legal/review boards when required. Store PII according to policy; de-identify synthesis artifacts.
Interview techniques
| Style | Structure | Pros | Cons |
|---|---|---|---|
| Structured | Fixed question order and wording | Comparable across sessions; easy for novices | Miss emergent themes |
| Semi-structured | Guide with core topics + probes | Balance consistency and depth | Needs skilled moderators |
| Unstructured | Minimal script; exploratory | Rich for novel domains | Hard to compare; time-consuming |
Probing: 5 Whys — chain “why” to root causes (watch for fatigue). Laddering — link features to consequences to values. Think-aloud — ask users to verbalize during tasks (evaluative); avoid leading the narrative.
Synthesis methods
- Affinity diagramming — cluster observations into themes bottom-up.
- Journey mapping — stages, touchpoints, emotions, pain points, opportunities.
- Persona development — behavioral archetypes from patterns (not stereotypes).
- Empathy mapping — says/thinks/does/feels for a scenario.
- Jobs-to-be-done — circumstances, struggles, desired outcomes (“hire” the product).
Insight synthesis funnel
Research repository
Democratize access so teams do not re-ask the same questions. Use a tagging taxonomy (topic, segment, product area, method, date). Atomic research stores bite-sized findings with evidence links so they compose into larger narratives. Tools such as Dovetail and EnjoyHQ support search, highlights, and insight libraries; spreadsheets work for small teams if discipline is maintained.
Usability testing specifics
| Dimension | Options | Notes |
|---|---|---|
| Moderation | Moderated vs unmoderated | Moderated: probes and clarifications. Unmoderated: scale and speed; write tasks very carefully |
| Setting | Remote vs in-person | Remote widens geography; in-person suits physical products or sensitive contexts |
| Tasks | Realistic goals, clear success criteria, avoid leading hints | Pilot tasks before full run |
| Protocol | Think-aloud | Explain once; stay neutral; note where users struggle |
| SUS | System Usability Scale — 10 items, 0–100 score | Compare over time or to industry benchmarks with caution |
Research deliverables
| Deliverable | Contents | Best for |
|---|---|---|
| Personas | Goals, behaviors, frustrations (evidence-backed) | Alignment, scenario writing |
| Journey maps | Stages, emotions, opportunities | Service design, prioritization |
| Empathy maps | Per scenario | Workshops, quick shared understanding |
| Insight reports | Questions, method, findings, limitations, recommendations | Stakeholders, decisions |
| Video highlights | Short clips with consent | Empathy building |
| Design principles | Durable heuristics from research | Critique and tradeoffs |
Anti-patterns
- Leading questions (“Don’t you think this is easier?”) — biases responses.
- Confirmation bias — only citing data that supports a preferred solution.
- Testing with colleagues — false confidence; not representative users.
- Research without action — no owner, no decision, no follow-up erodes trust.
External references
- Just Enough Research — Erika Hall (pragmatic research for teams).
- NN/g: User Research Methods — methods encyclopedia and courses.
- Interviewing Users — Steve Portigal (moderation and field practice).
Keep project-specific accessibility audits in docs/product/ and remediation plans in docs/development/, not in this file.