Purpose
Turn a loosely defined problem space into a durable research framework — a structured set of research questions, investigative lenses, and method mappings that a team can use to run coherent, cumulative research over months. Solves the "every study starts from scratch" problem and gives research a defensible structure product and design can plan against.
When to Use
Invoke this skill when the user asks to:
- Stand up a research program or agenda for a new product area or problem space
- Convert scattered, one-off study requests into a coherent research strategy
- Build a reusable structure or set of lenses for ongoing product discovery
- Define the research questions and dimensions that should guide a quarter or year of research
- Create a framework a research team can share with product and design stakeholders
- Systematize how a team decides what to research next and why
Inputs Required
Before starting, collect:
- Problem space: 2–4 sentence description of the domain being investigated (e.g., onboarding activation, enterprise admin workflows, churn)
- Product area: The specific product, feature set, or surface this framework governs
- Primary users/segments: Who is being researched, including distinct roles or tiers
- Business objective: The outcome this research is meant to inform (activation lift, expansion, retention, roadmap prioritization)
- Known unknowns: What the team currently does not understand or keeps arguing about
- Existing evidence: Prior research, analytics signals, or support/sales themes already available
- Time horizon: How long this framework should stay useful (quarter, half, year)
If known unknowns or existing evidence are missing, infer likely gaps from the problem space and flag them as assumptions to validate.
Instructions
Step 1: Define the Framing Question
Compress the problem space into one durable framing question — the north-star inquiry the whole framework serves. It should be broad enough to hold a year of research but specific enough to be answerable.
- Format: "How do [segment] [attempt to achieve outcome] within [product area], and what determines success or friction?"
- Reject framing questions that are actually feature validation questions — those belong to individual studies, not the framework
- Note the business objective this framing question ladders up to
Step 2: Construct 4–6 Research Lenses
Lenses are the stable dimensions through which the problem space is examined. They persist even as individual studies rotate. Build lenses from the set below, selecting and adapting to the domain — do not force all six if fewer fit cleanly.
Lens A — Behavioral: What users actually do — workflows, sequences, workarounds, abandonment points.
Lens B — Motivational: Why they do it — goals, jobs-to-be-done, triggers, and the outcome they are hired to deliver.
Lens C — Contextual: The conditions around the behavior — team structure, tooling, constraints, environment, integration surface.
Lens D — Perceptual: How users interpret the product — mental models, expectations, points of confusion, trust.
Lens E — Temporal: How behavior and needs change over time — first-run vs. tenured, seasonal, lifecycle stage.
Lens F — Comparative: How the current experience measures against alternatives — competitors, prior tools, manual processes.
For each selected lens, write:
- A one-line definition scoped to this problem space
- 2–4 core research questions expressed in the user's terms, not the product's
- The primary signal type it produces (attitudinal vs. behavioral; qualitative vs. quantitative)
Step 3: Map Methods to Lenses
For each lens, recommend the method(s) that produce valid evidence for its questions.
- Behavioral / Temporal → usage analytics, session replay, diary studies, longitudinal panels
- Motivational / Perceptual → in-depth interviews, JTBD interviews, concept tests, card sorts
- Contextual → contextual inquiry, stakeholder interviews, workflow mapping
- Comparative → competitive teardowns, switch/win-loss interviews, benchmarking surveys Note the minimum viable sample or data volume for each method, and flag which methods the team can run now vs. which require new instrumentation.
Step 4: Build Hypothesis Scaffolds
For each lens, draft 1–2 falsifiable hypotheses in the form: "We believe [segment] [does/experiences X] because [driver], which we'll know is true if [observable evidence]."
- Hypotheses must be disprovable — if no observation could refute it, rewrite it
- Tie each hypothesis to the lens and method that would test it
Step 5: Define Evidence Thresholds
Specify what counts as "enough" to act on a finding, so the team doesn't over- or under-research.
- Directional (proceed with caution): pattern in 3–5 qualitative sessions or a single unvalidated quant signal
- Confident (act): convergence across two method types, or a quant signal at a defined significance level plus qualitative corroboration
- Settled (foundational belief): repeated across studies and stable over the time horizon State which decisions each threshold is sufficient for.
Step 6: Build the Prioritization Matrix
Rank the research questions across all lenses on two axes:
- Decision impact: How much a clear answer changes a roadmap, design, or investment decision (High / Med / Low)
- Current uncertainty: How little the team currently knows (High / Med / Low) Sequence the framework's research agenda: High impact + High uncertainty first. Output an ordered list of the first 3–5 studies the team should run, each tagged with its lens, method, and the decision it unblocks.
Step 7: Assemble the Framework Document
Compile everything into a single reusable artifact that a research manager can share, defend in planning, and revisit each quarter.
Output Format
- Structure: A single framework document with clearly delimited sections; lenses presented in a consistent repeating block; prioritized agenda as an ordered list or table
- Length: 1,100–1,500 words
- Format: Structured document — headers, bullets, and one prioritization table; not prose-heavy
- Sections: Framing Question → Research Lenses (4–6 repeating blocks) → Method Map → Hypothesis Scaffolds → Evidence Thresholds → Prioritization Matrix → First 3–5 Studies (sequenced)
- Tone: Research strategist to research team — rigorous, plainly worded, opinionated about sequencing; no methodology jargon without a plain-language gloss
Quality Checklist
Before delivering output, verify:
- [ ] Exactly one framing question is stated and it is not a disguised feature-validation question
- [ ] Between 4 and 6 lenses are defined, each with a definition, 2–4 questions, and a signal type
- [ ] Every lens has at least one mapped method with a stated minimum sample or data volume
- [ ] Every hypothesis is falsifiable — each includes an observable "we'll know it's true if" condition
- [ ] The prioritization matrix scores questions on both impact and uncertainty, and the first 3–5 studies are sequenced accordingly
- [ ] All questions are written in the user's terms, not the product's feature names
- [ ] Any missing input (unknowns, prior evidence) is surfaced as a flagged assumption
Notes
- Lenses are the durable asset — studies rotate, but a well-built lens set stays useful for a full time horizon. Resist collapsing lenses into individual research questions.
- The most common failure mode is a framework that's really a backlog of one-off study requests. If the output reads like a to-do list, return to Step 2 and rebuild the structural dimensions.
- Contextual and Comparative lenses are the ones teams most often skip and most often need — include them unless the domain clearly excludes them.
- Recommend revisiting the framework at each quarter boundary: retire settled beliefs, re-score uncertainty, and re-sequence the agenda.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills