Step-by-Step Implementation Guide

graph LR
    P1["📋 PHASE 1<br/>Research Planning<br/>15-20 min"]
    S1["Step 1:<br/>Define Problem<br/>Scope"]
    S2["Step 2:<br/>Set Research<br/>Objectives"]
    S3["Step 3:<br/>Prepare<br/>Documentation"]
    P2["🤖 PHASE 2<br/>AI-Assisted Research<br/>30-40 min"]
    S4["Step 4:<br/>Customize<br/>MCP Prompt"]
    S5["Step 5:<br/>Execute AI<br/>Research"]
    S6["Step 6:<br/>Quality<br/>Verification"]
    P3["🌳 PHASE 3<br/>Tree Construction<br/>25-35 min"]
    S7["Step 7:<br/>Extract Core<br/>Problem"]
    S8["Step 8:<br/>Map Root<br/>Causes"]
    S9["Step 9:<br/>Identify<br/>Effects"]
    P4["💬 PHASE 4<br/>Stakeholder Prep<br/>15-20 min"]
    S10["Step 10:<br/>Convert to<br/>Questions"]
    S11["Step 11:<br/>Plan Validation<br/>Approach"]

    P1 --> S1 --> S2 --> S3 --> P2
    P2 --> S4 --> S5 --> S6 --> P3
    P3 --> S7 --> S8 --> S9 --> P4
    P4 --> S10 --> S11

    style P1 fill:#F59E0B,stroke:#D97706,stroke-width:2px,color:#1F2937
    style P2 fill:#F59E0B,stroke:#D97706,stroke-width:2px,color:#1F2937
    style P3 fill:#F59E0B,stroke:#D97706,stroke-width:2px,color:#1F2937
    style P4 fill:#F59E0B,stroke:#D97706,stroke-width:2px,color:#1F2937
    style S1 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S2 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S3 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S4 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S5 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S6 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S7 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S8 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S9 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S10 fill:#72B043,stroke:#5A8F36,stroke-width:1px,color:#fff
    style S11 fill:#10B981,stroke:#059669,stroke-width:2px,color:#fff

Follow each phase and step below for detailed instructions:

Phase 1: Research Planning (15-20 minutes)

Invest time upfront to clarify your scope and objectives. This prevents wasted effort later.

Step 1: Define Your Problem Scope

What to do:

• Write a clear 1-2 sentence problem statement
• Specify affected population and geographic area
• Note any time constraints or urgency factors
• List what you already know (even if limited)

Example output:

Problem: Young adults aged 18-25 in rural Nyanza region, Kenya, have limited access to decent, stable employment opportunities.

Population: ~50,000 youth in 5 counties; mixed gender; most secondary school graduates.

What we know: National youth unemployment ~20%; agriculture declining; anecdotal skills-job mismatch.

Step 2: Set Research Objectives

Answer these questions:

• What do you need to understand about causes?
• What effects/consequences are you curious about?
• What would help you design better interventions?
• What stakeholders do you eventually need to engage?

Step 3: Prepare Documentation System

Set up your workspace:

• Create a folder for sources and citations
• Start a document with Problem Tree template (see Templates & Tools page)
• Create a tracking system for evidence (E) vs. assumptions (A)
• Have a place ready to save AI outputs

Phase 2: AI-Assisted Research (30-40 minutes)

Use MCP to accelerate desk review while maintaining rigor.

Step 4: Customize Your MCP Prompt

What to do:

• Copy the complete MCP template from the Model Context Protocol page
• Replace all {{placeholders}} with your specific context from Step 1
• Review each block for relevance and clarity
• Prepare to copy-paste into ChatGPT, Claude, or your preferred AI tool

Time estimate: 10 minutes

Step 5: Execute and Review AI Research

What to do:

• Paste your complete MCP prompt into AI tool
• Wait for output (usually 1-2 minutes)
• Copy and save all outputs to your documentation folder immediately
• Do initial scan for obvious errors, irrelevance, or red flags

Time estimate: 5 minutes

Step 6: Quality Verification

What to do:

• Check credibility of 3-5 sources by opening links or searching for publications
• Verify key statistics against original sources if possible
• Note any contradictions or gaps in coverage
• Flag items that seem like assumptions vs. evidence
• If sources don't work or seem suspicious, mark those findings (A)

Time estimate: 15-20 minutes

Phase 3: Problem Tree Construction (25-35 minutes)

Organize AI findings into the three-part Problem Tree structure.

Step 7: Extract and Organize Core Problem

What to do:

• Refine problem statement based on research insights
• Ensure it specifies who, what, where (population, issue, geography)
• Remove any causes or solutions that crept into the statement
• Confirm it's observable, measurable, and urgent

Check your work: Can you verify this problem exists through data or direct observation? Does it avoid embedding causes or solutions?

Time estimate: 5 minutes

Step 8: Map Root Causes by Levels

What to do: Organize causes into 3 levels of depth:

• Level 1 (Direct causes): What immediately creates the problem? These are the most visible causes.
• Level 2 (Underlying causes): Why do the direct causes exist? Go deeper with "Why?" questioning.
• Level 3 (Structural causes): What systems, policies, or conditions enable underlying causes? This is the deepest level.

Tag each cause:

• (E) for evidence-based: Supported by credible research you verified
• (A) for assumption: Logical but needs validation with stakeholders

Example structure:

Time estimate: 10-15 minutes

Step 9: Identify Effects by Time Horizon

What to do: Map consequences across time horizons:

• Immediate effects (0-6 months): What happens right now because of this problem?
• Medium-term effects (6 months-2 years): What develops over time?
• Long-term effects (2+ years): What are lasting consequences?

Consider different levels of impact:

• Individual: Direct impact on affected people
• Family/household: Ripple effects on families
• Community: Broader social and economic consequences
• System: Policy, institutional, or structural impacts

Tag each effect (E) or (A) just like causes.

Time estimate: 10-15 minutes

Phase 4: Stakeholder Preparation (15-20 minutes)

Convert your assumptions into questions that guide meaningful community engagement.

Step 10: Convert Assumptions to Questions

What to do:

• Take every item marked (A) and turn it into an open-ended question
• Avoid leading questions that push toward your preferred answer
• Focus on understanding experience, not confirming theories
• Aim for 8-10 core questions that cover major assumptions

Examples of good vs. bad questions:

Time estimate: 10 minutes

Step 11: Plan Validation Approach

What to do:

• Identify which stakeholder types could answer your questions (youth, employers, parents, community leaders, government officials)
• Consider how you'll reach and engage them meaningfully (focus groups, interviews, community meetings)
• Set realistic timeline for validation conversations
• Prepare to update your Problem Tree based on what you learn

Important mindset: Go to stakeholders ready to be surprised, challenged, and corrected. The goal isn't to confirm what AI told you—it's to learn what AI missed, got wrong, or oversimplified.

Time estimate: 5-10 minutes

Quick Reference Checklist

Use this checklist to track your progress:

Common Questions

"What if AI gives me generic results in Phase 2?"

Go back to Step 4 and add more specific geographic/population details in your Knowledge Context block. Follow up with: "Focus specifically on [your region] and provide more context-specific causes."

"How do I know if I've gone deep enough in Step 8?"

You've gone deep enough when asking "Why?" one more time would take you beyond what your project can feasibly address. Stop at actionable causes.

"Can I skip Phase 2 and just do traditional research?"

Absolutely! MCP is optional. But if you use it, you'll likely save 4-6 hours of manual literature review while achieving similar quality.

Ready to See Examples?

Now that you understand the process, explore Templates & Tools for ready-to-use templates and Examples to see completed Problem Trees from real projects.