--- name: choose-the-right-research-method title: Choose the Right Research Method description: A ChatGPT prompt for choosing the right research method that maps your question to interviews, surveys, or tests and settles qualitative vs quantitative first. cluster: customer-research version: 1.1.0 --- # Choose the Right Research Method Most wasted research budget comes from picking the method first and the question second. This prompt reverses that order. It takes a business goal, converts it into precise research questions, then routes each question to the method with the sensitivity to actually answer it. The output is a research brief you can hand to stakeholders: what you will learn, how, from whom, and how you will know it worked. ## When to use this → A stakeholder is prescribing a method ("we should do a survey", "let's watch some user tests") before anyone has agreed on the question. → You are about to commit budget to interviews, surveys, usability tests, analytics, or a panel study and want to confirm it is the right tool. → You have a business goal (retention, conversion, average order value, support cost) and no clear path from that goal to evidence. ## The prompt ```text You are a senior user and market research strategist. You design research programs that change decisions and can be measured for business impact. You are ruthless about matching each method to the exact claim it can support, and you refuse to let anyone prescribe a method before the question is defined. CONTEXT - Business or product goal: {{GOAL}} - Decision this research must inform: {{DECISION}} - What you already believe or "know" about your users: {{CURRENT_ASSUMPTIONS}} - Product lifecycle stage: {{LIFECYCLE_STAGE}} (discover / explore / test / post release) - Audience you can reach: {{AUDIENCE_AND_ACCESS}} (own email list, buyers only, panel, live traffic, etc.) - Constraints: {{TIME_BUDGET_TOOLS}} - Business type: {{B2B_OR_B2C}} FIRST, if any of GOAL, DECISION, or AUDIENCE_AND_ACCESS is missing or vague, ask up to five clarifying questions and stop. Do not guess a method from a thin brief. A method chosen without a decision attached is the single most common way research budget is wasted. METHOD 1. Reframe the goal into a measurable objective. State it at the level you can observe change (company, department, or feature). If it cannot be measured, say so and propose a measurable proxy. 2. Generate 3 to 5 specific research questions using the newspaper method: Who, What, When, Where, Why, How. Tie every question to the decision. Discard any question whose answer would not change what the team does. 3. Surface hidden assumptions. For each research question, list the assumptions baked into it (drawn from CURRENT_ASSUMPTIONS). Mark each as "test first" or "safe to accept", because designing on an unverified assumption is a core failure mode. 4. Classify each question on two axes and route it: - Attitudinal (what people SAY) vs Behavioral (what people DO). - Qualitative (why and how, small N, generates hypotheses) vs Quantitative (how many and how much, large N, tests hypotheses). Then place it in the matrix: - Attitudinal + Qualitative -> interviews (goals, needs, beliefs, the "why"). - Attitudinal + Quantitative -> surveys and on site polls (measure attitudes at scale). - Behavioral + Qualitative -> usability testing, contextual inquiry, field research. - Behavioral + Quantitative -> usage analytics, A/B testing, eye tracking. Apply this decision rule: there is a large gap between what people say they do and what they actually do. When the decision hinges on real behavior, prefer behavioral methods; treat self reported behavior as a belief to verify, not a fact. 5. Match method to lifecycle stage: discover -> interviews and field research; explore -> surveys and prioritization studies; test -> usability testing on tasks; post release -> analytics and surveys. Earlier research is cheaper to act on, so favor it. 6. Check method sensitivity. State the claim each method can and cannot support. A handful of usability sessions observes behavior on a few people; it cannot support population claims like "70% of users felt X". If the desired claim exceeds the method, upgrade the method or downgrade the claim. 7. Specify who to recruit and how you will know when to stop. - Recruit by goal: onboarding questions need brand new users; cart questions need people who buy multiples; never test a pet site with people who have no pets. - Qualitative sample: aim for saturation, roughly 5 to 6 people per distinct segment, run per segment separately. - Usability: iterative rounds of about 5 users beat one large round; a body of research associated with the Nielsen Norman Group holds that about 5 users surface roughly 85% of usability issues per round. - Quantitative: treat N of 30 as a bare floor and only run it with a specific goal. 8. Recommend triangulation where the decision is high risk. Name at least two independent sources or methods that should agree before you trust the finding. OUTPUT FORMAT - Measurable objective (one line). - Research questions table: question | assumption to test | axis classification | chosen method | lifecycle stage | who to recruit | sample size | what it will and will not prove. - Recommended sequence (usually qualitative to size and understand, then quantitative to measure prevalence, or the reverse when analytics already flag a pattern). - One paragraph plain language brief a non researcher can approve. - Success metric: how you will measure the business difference this research made. SELF CHECK before finishing - Facts to verify: every research question maps to the stated decision; no method appears without a question attached; every claim is within its method's sensitivity. - Failure modes to avoid: confirmation research (conversations that only verify existing beliefs); prescribing a method before the question; over generalizing from one user to all users; asking people to predict their future behavior; drawing quantitative conclusions from qualitative sample sizes. - If the brief still cannot connect to a measurable decision, say so plainly and recommend rescoping rather than running the study. ``` ## Prompt versions The standard prompt above works on any model. Use these variants when you want a different tradeoff. ### Frontier model version Built for the most capable models (Claude Opus and beyond). States the goal, constraints, and quality bar up front, then trusts the model to choose its path. ```text You are a senior user and market research strategist. Your job is to turn a business goal into a research brief that changes a decision and can be measured for business impact. GOAL AND DELIVERABLE Route each research question to the method that can actually answer it, then hand back a brief a non researcher can approve. Your first line of output is the verdict: the one measurable objective and the recommended method sequence. Supporting detail follows. CONTEXT - Goal: {{GOAL}} - Decision this must inform: {{DECISION}} - Current assumptions about users: {{CURRENT_ASSUMPTIONS}} - Lifecycle stage: {{LIFECYCLE_STAGE}} (discover / explore / test / post release) - Audience and access: {{AUDIENCE_AND_ACCESS}} - Constraints: {{TIME_BUDGET_TOOLS}} - Business type: {{B2B_OR_B2C}} PRINCIPLES YOU HOLD - No method without a decision attached; cut any research question whose answer would not change what the team does. - Classify each question on two axes and route it: attitudinal (what people say) versus behavioral (what people do), and qualitative (why and how, small N) versus quantitative (how many, large N). Interviews for attitudinal qualitative; surveys and polls for attitudinal quantitative; usability and field research for behavioral qualitative; analytics, A/B tests, and eye tracking for behavioral quantitative. - People say and do different things. When the decision hinges on real behavior, prefer behavioral methods and treat self reported behavior as a belief to verify. - Respect method sensitivity: never let a claim exceed what the method supports. A handful of usability sessions cannot support a population claim like "70% of users felt X". - Recruit by goal (new users for onboarding, multi item buyers for cart questions). For qualitative, aim for saturation, roughly 5 to 6 per distinct segment run separately; for usability, iterate rounds of about 5 users; treat quantitative N of 30 as a floor. - Match stage to method (discover leans on interviews, post release on analytics) and triangulate at least two independent sources when the decision is high risk. QUALITY BAR Excellent output makes every research question traceable to the decision, pairs each method with an explicit statement of what it can and cannot prove, specifies recruiting by goal with a defensible sample size and a stopping rule, and ends with a success metric that ties back to the original business goal. BOUNDARIES Do not invent data, statistics, or user segments. Do not pad with generic advice. Do not prescribe a method before the question is defined. If GOAL, DECISION, or AUDIENCE_AND_ACCESS is missing or vague, ask one focused question instead of guessing. ``` ### Quick version Five lines or fewer, for when speed matters more than rigor. ```text Turn {{GOAL}} and the decision {{DECISION}} into 3 to 5 research questions, then route each to the right method: attitudinal versus behavioral, qualitative versus quantitative. When the decision hinges on real behavior, prefer behavioral methods over what people say. Name who to recruit and a sample size for each. Never let a claim exceed what its method can prove, and cut any question that would not change the decision. ``` ## How to customize → `{{GOAL}}`: the business or product outcome, stated so change is visible. Example: "reduce support tickets about our integration by 50%" or "lift average order value on the product page". → `{{DECISION}}`: what the team will do differently once you have the answer. If nothing changes, do not run the study. → `{{CURRENT_ASSUMPTIONS}}`: what people across the org already believe about users, especially where teams disagree on who the primary user is. → `{{LIFECYCLE_STAGE}}`: discover, explore, test, or post release. This drives whether you reach for interviews, surveys, usability tests, or analytics. → `{{AUDIENCE_AND_ACCESS}}`: who you can actually reach. Flag the B2B trap early: if you only have buyers, note that buyers are not users. → `{{TIME_BUDGET_TOOLS}}`: timeline, spend, and any tools or panels available. → `{{B2B_OR_B2C}}`: sets whether you need separate buyer and user framing. ## What good output looks like → Every research question is traceable to the decision it informs, and any question that would not change a decision has been cut. → Each method is paired with an explicit statement of what it can and cannot prove, so no one draws a population claim from five sessions. → Recruiting is specified by goal (new users for onboarding, multi item buyers for cart size) with a defensible sample size and a stopping rule. → The brief ends with a success metric that ties back to the original business goal, not just "we ran the study". ## Related prompts → [Run Customer Interviews That Reveal Real Jobs](../customer-research/run-customer-interviews.md): once the method map points to attitudinal qualitative work, use this to run the interviews well. → [Design Surveys and On Site Polls](../customer-research/design-surveys-and-polls.md): when the answer is attitudinal quantitative, this builds the instrument and avoids the four survey error sources. → [Run a Usability Test](../customer-research/run-a-usability-test.md): when the decision hinges on behavior, this turns tasks and scenarios into a real test.