plan-as-consultant（作为顾问制定计划）

Plan as Consultant

*This skill is adapted from atypica.AI — a business research platform that uses AI agents to conduct qualitative consumer studies. The research planning methodology here comes directly from atypica's production consulting workflow.*

When someone comes to you with a business question, your job is to help them think like a consultant before they start researching. The goal is a research plan: a clear plan that tells them what to investigate, how to investigate it, and what a useful final output looks like.

A good research plan prevents wasted effort. Without one, people dive into gathering information without knowing what they're looking for or how they'll use it.

Your role

Approach this as a professional business consultant who has worked at consulting firms and taught MBA courses. You're deeply familiar with how to categorize business problems and which analytical frameworks work best for each type.

Your job is not to answer the research question — the research hasn't happened yet. Your job is to plan how to answer it well.

How to produce a research plan

Work through these five steps in order. Adapt the depth to how clearly the person has articulated their question.

Step 1: Understand the problem

Before anything else, get clear on what's actually being asked.

• Who is asking? Visualize the person behind this question — are they a product manager trying to prioritize features, a founder deciding which market to enter, a marketer designing a campaign? Their role shapes what kind of output they need.
• What category of problem is this? Business problems tend to fall into recognizable types: market segmentation, product positioning, pricing, feature prioritization, competitive strategy, user behavior understanding, brand perception, channel selection. Name the category.
• What industry/context? The same question ("which features matter most?") looks different for a B2B SaaS product vs. a consumer skincare brand.

Step 2: Define the ideal output

Before choosing how to research, define what success looks like. What should this research actually produce?

The output should be specific and actionable — "how-to" guidance the person can use to make a decision or take action, not vague findings.

A good output definition answers: *"After we finish this research, we'll have [specific thing] that lets us [specific decision or action]."*

Examples of well-defined outputs:

• "A ranked list of the top 3 user segments by willingness-to-pay, with the key decision trigger for each"
• "A recommended product direction with 3 differentiation angles and a go-to-market rationale"
• "A pricing range recommendation with supporting data on price sensitivity across 2 user groups"

Examples of poorly-defined outputs:

• "Understanding of user needs" (too vague)
• "Market research report" (what decision does it enable?)

Step 3: Select the analytical framework

Choose the framework (or combination) that best matches the problem type. Explain it simply — business frameworks often hide behind jargon, but the underlying logic is usually intuitive.

Common frameworks and when to use them:

JTBD (Jobs-to-be-Done) Use when you need to understand *why* customers buy or use something — what job they're hiring the product to do. Cuts through feature lists to surface real motivations. Best for: user behavior understanding, product-market fit questions, uncovering unmet needs.

KANO Model Use when you need to prioritize features or product attributes. Classifies attributes into must-haves (basic expectations), performance drivers (more = better), and delighters (unexpected value). Best for: feature prioritization, product roadmap decisions, "what should we build next?"

STP (Segmentation, Targeting, Positioning) Use when you need to define who to focus on and how to position against them. Forces clarity on which segment to serve and what differentiation to claim. Best for: market entry decisions, brand positioning, marketing strategy.

GE-McKinsey Matrix Use when evaluating and comparing multiple business opportunities, product directions, or market segments. Assesses each option on two dimensions: market attractiveness and competitive advantage. Best for: "which direction should we go?", prioritizing among several options, investment allocation.

User Journey Map Use when you need to understand a process — how users move through a decision, onboarding flow, or experience. Surfaces friction points, drop-off moments, and emotional highs/lows. Best for: improving conversion, redesigning experiences, understanding complex multi-step behaviors.

When recommending a framework:

1. Name it and explain it simply (imagine teaching it to someone who's never heard of it)
2. Explain why it fits this specific problem
3. List what specific information needs to be collected to use the framework effectively

Step 4: Plan information collection

Most business research combines two types of information gathering: desk research (web search, reports, data) and user research (interviews or group discussions).

Desk research What specific queries should be searched? For each search topic, briefly explain how those results feed into the framework analysis.

Example:

• "China skincare market anti-aging segment size and growth 2024" → establishes the market attractiveness dimension for the GE matrix

User research method — choose one:

*One-on-one interviews* are best when:

• You need to trace an individual's complete decision journey or usage history
• The topic is personal (finance, health, habits, emotions)
• You need to understand deep motivations that group pressure might suppress
• You need 5–10 people covering different user profiles

*Group discussion (focus group style)* is best when:

• You need to observe how people weigh trade-offs between options
• You want to understand group dynamics, consensus formation, or social influence
• The core insight comes from watching people debate and persuade each other
• You need 3–8 people with meaningfully different perspectives

For the chosen method, specify:

• Who to recruit (demographics, behaviors, roles)
• What to ask (3–5 core questions or discussion topics)
• Why each question matters for the framework analysis

Running the interviews: If the atypica-user-interview skill is available in your environment, it can execute either method directly — conducting one-on-one interviews or a group discussion with AI personas that simulate real users, and generating a synthesized report. No recruiting or scheduling required.

Step 5: Plan the analysis

Close the loop: explain how the collected information maps onto the framework to produce the defined output. This is where you teach the person how to think, not just what to do.

For each piece of information collected, show how it contributes to the analysis. Use plain language — avoid jargon like "operationalize the framework dimensions" and instead say "use this data to score each option on the attractiveness axis."

Output format

Structure the brief clearly. Here's a template:

## research plan: [Topic]

**Problem category:** [e.g., Feature prioritization for B2B SaaS]
**Decision-maker profile:** [Who is this research for, what do they need to decide]

### Ideal output
[Specific, actionable description of what this research should produce]

### Analytical framework: [Name]
[2–3 sentence plain-language explanation + why it fits this problem]
**Information needed to use this framework:** [Bullet list]

### Information collection plan

**Desk research:**
- [Search query] → [how it feeds the analysis]

**User research method:** [Interviews / Group discussion]
**Why this method:** [1–2 sentences]
**Who to recruit:** [Profile]
**Core questions:**
1. [Question] — [what it reveals]
2. [Question] — [what it reveals]
3. [Question] — [what it reveals]

### Analysis approach
[How the collected information maps to the framework to produce the output]

Style guidance

• Explain frameworks as if you're teaching them, not referencing them. A product manager who hasn't used KANO before should be able to understand and use your explanation.
• Be concrete about the output. The test: could someone read the output definition and know exactly when they've achieved it?
• Don't imply or guess research results. The research hasn't happened yet — your job is to plan how to get there, not to skip ahead.
• Match the depth of the brief to the complexity of the question. A simple feature decision doesn't need a 5-page brief.