How to Write a Research Brief: Templates, Examples, and AI-Assisted Generation

A research brief is the foundation of effective user interviews. Before you write a single question, a clear brief answers three critical questions: What decision does this research inform? Who should we talk to? How should we approach the conversation?

Most interview problems — vague findings, rambling sessions, insights that don't drive decisions — trace back to a weak or missing research brief. This guide shows you how to write a brief that produces interview data you can actually act on, and how Koji's AI can generate one automatically when you need to move fast.

What Is a Research Brief?

A research brief is a structured document that defines the purpose, scope, and approach of a user research project before fieldwork begins. It serves two functions:

Alignment: It ensures the researcher, stakeholders, and any AI interviewer all share the same understanding of what the research is trying to achieve.

Interview quality: The brief is the context that enables intelligent follow-up questions. Without a brief, AI probing is generic. With a specific brief, every follow-up targets the exact behaviors and decisions you're studying.

In Koji, your research brief feeds directly into the AI interviewer. The more specific your brief, the more targeted and insightful the AI's probing becomes.

The 7 Components of an Effective Research Brief

1. Problem Statement

A one-to-two sentence description of the specific problem or opportunity you're researching.

Weak: "We want to understand why users aren't adopting the new feature."

Strong: "Users who see the new analytics dashboard within 7 days of signup are not returning to it after the first visit — we want to understand what prevents repeated use."

The difference: the strong version names the specific behavior gap and implies a testable hypothesis.

2. Decision to Inform

What action or decision will this research support? Being explicit forces you to confirm the research is decision-relevant — not just interesting.

Examples:

• "Whether to simplify the onboarding flow or add contextual help tooltips"
• "Whether to build a native mobile app or optimize the web experience for mobile"
• "Which of three pricing tier structures to lead with in outbound sales"

If you can't name a specific decision, the research is probably too exploratory to scope properly. Run a discovery sprint first, then write a more targeted brief.

3. Current Hypothesis

What do you currently believe to be true? Writing your hypothesis before the research forces intellectual honesty — you can't pretend you didn't have a prior belief after you've seen the data.

Example: "We believe users find the analytics dashboard overwhelming because it displays too many metrics by default, with no obvious starting point."

Your hypothesis might be right, wrong, or partially right. The research will tell you. But writing it prevents confirmation bias — you're committing to a prediction that can be tested.

4. Success Criteria

What would you see in the data that would tell you the research succeeded?

Example: "We'll understand the top 3 specific barriers to repeated dashboard use, with supporting participant quotes and at least 6 participants articulating each barrier clearly."

Success criteria also tell you when to stop collecting data. If your criterion is data saturation, you stop when new interviews add no new themes.

5. Target Participant Profile

Who should you talk to? The critical mistake here is defining participants demographically rather than behaviorally.

Demographic definition (weak): "Marketing managers, 25–45 years old, at mid-size B2B companies"

Behavioral definition (strong): "Someone who has logged into the analytics dashboard at least once in the past 30 days but has not returned in the past 2 weeks, and is the primary person responsible for reporting metrics to leadership"

The behavioral definition produces a sample that has actually experienced the specific problem you're studying. Demographics alone don't guarantee that.

Also define a screening question — a single yes/no or short-answer question that validates participant fit before they begin the interview.

6. Methodology Framework

Which research methodology are you following? The methodology determines how the AI interviewer approaches the conversation: which question patterns to use, how to probe, and what to avoid.

Common methodologies supported by Koji:

• The Mom Test (Rob Fitzpatrick): Talk about their life, not your idea. Ask about past behavior, not hypotheticals. Dig for specifics. Best for early-stage product discovery.
• Jobs to Be Done: Focus on the progress the participant is trying to make. Understand what pushed them away from the old approach and what pulled them toward something new. Best for understanding switching behavior and unmet needs.
• Customer Discovery (Steve Blank): Validate that the problem exists and is painful enough to motivate action. Understand current alternatives and their shortcomings. Best for pre-product or new-market research.
• Exploratory / Generative: Follow the participant's energy. Let them lead. Best for truly open discovery where you don't yet know what you don't know.

In Koji, selecting a methodology automatically configures the AI interviewer with the right probing principles, question patterns, and anti-patterns for that framework.

7. Interview Guide (Questions)

The research brief culminates in the interview guide. Write 10–15 questions, design to cover 6–8 in 45 minutes. For each question, specify:

• The question text (open-ended, past-behavior framing)
• The question type (open_ended, scale, single_choice, multiple_choice, ranking, or yes_no)
• Probing depth — how many follow-ups should the AI pursue on this question?
• Any specific probing instructions ("When they mention a workaround, ask how often they use it")

Using Koji's six structured question types lets you collect both qualitative depth (from open-ended questions with AI follow-ups) and quantitative data (from scale, choice, and ranking questions) in a single interview — giving you both the story and the numbers.

How Long Should a Research Brief Be?

The right length is "as short as possible while covering all 7 components." A good brief for a focused 15-interview study is typically 400–600 words — specific enough to guide intelligent probing, short enough to write and read quickly.

Avoid letting your brief become background reading or a literature review. Every sentence should directly improve how the interview is conducted.

How Koji Generates Research Briefs Automatically

If writing a detailed research brief feels like a barrier, Koji's AI can generate one for you. Describe your research goal in plain language — one or two sentences — and Koji drafts:

• A specific problem statement
• A hypothesis to test
• A behavioral participant profile with screening question
• A recommended methodology
• A complete interview guide with question types, probing depth, and follow-up instructions

You review and edit each component before deploying the study. The AI-generated brief is a starting point, not a constraint — and it typically takes under 5 minutes to generate and refine.

This feature is especially useful for teams new to user research, or researchers who need to move quickly without spending hours on brief preparation.

Research Brief Template

Use this as a starting point:

---

Problem Statement: [1–2 sentences describing the specific behavior gap or problem]

Decision to Inform: [The specific decision this research will support]

Current Hypothesis: [What we currently believe to be true]

Success Criteria: [What data would confirm the research succeeded]

Target Participant:

• Required experience: [What must they have done/experienced?]
• Behavior of interest: [What behavior are we studying?]
• Screening question: [The one question that confirms fit]

Methodology: [Mom Test / Jobs to Be Done / Customer Discovery / Exploratory]

Research Questions:

1. [Question text] — Type: open_ended, Probing: 2 follow-ups
2. [Question text] — Type: scale 1–10, Probing: anchor on low scores
3. [Question text] — Type: single_choice, Options: A / B / C ...

---

Common Research Brief Mistakes

Skipping the hypothesis: If you don't write what you currently believe before the research, you'll unconsciously interpret data to confirm it. Write it down, test it against the findings, and report honestly.

Demographic participant targeting: "Marketing managers" is a job title, not a behavioral profile. Define participants by what they've done, not who they are demographically.

Question-first thinking: Writing questions before defining your research goals means questions drift toward interesting-but-not-actionable. Always start with the decision to inform.

Too many questions: 15 questions in 45 minutes means 3 minutes per question — no room for probing. Design for 6–8 core questions with real follow-up depth.

No success criteria: Without criteria for success, you'll always feel like you need one more interview. Define when you have enough data before you start collecting it.

Vague methodology: "Semi-structured interviews" is not a methodology — it's a format. Choose a framework with specific principles (Mom Test, JTBD, Customer Discovery) that guide how the AI probes and what it avoids.

Brief Quality Checklist

Before deploying your study, verify your brief answers:

• What specific decision does this research inform?
• What is our current hypothesis (stated before seeing data)?
• What must participants have done to qualify?
• What methodology framework will guide probing?
• Are questions framed around past behavior, not hypotheticals?
• Does each question have a type and probing depth assigned?
• What would "enough data" look like?

Related Resources

• Structured Questions Guide: Mixing Qualitative and Quantitative in One Interview
• How to Conduct User Interviews: The Complete Guide
• Customer Discovery Interviews: Validate Assumptions Before Building
• Jobs to Be Done Framework: Understanding What Users Actually Want
• Thematic Analysis: Finding Themes in Interview Data
• AI Voice Interviews: Automated Voice User Research