analysis-results-collector skill

Analysis Results Collector

This skill helps you transform completed analytical work into clear, actionable documentation by conducting structured conversations that extract findings and build comprehensive results documents.

When to Use This Skill

Use this skill when:

• User has completed an analysis and needs to document results
• User mentions finishing hypothesis testing or investigation work
• User has an analysis plan (from analysis-planner or similar) and completed analyses
• User needs to create executive summaries or findings reports
• User asks to "document findings," "write up results," or "create analysis report"

Overview of Process

The skill guides users through a 4-phase conversation:

1. Hypothesis Triage - Identify which hypotheses were analyzed vs not analyzed
2. Validation Status - Determine which were validated, invalidated, or partially validated
3. Detailed Findings - Gather communication-ready summaries for each hypothesis
4. Recommendations - Capture prioritized, actionable next steps

After gathering information, generate a structured markdown document using the template.

Step 1: Parse the Analysis Plan

Action: Read the uploaded analysis plan to understand the structure and hypotheses.

Extract from the plan:

• Problem statement and North Star Metric
• List of all hypotheses (with their IDs like H1, H2, etc.)
• Hypothesis tier structure (Tier 1, 2, 3)
• Original timeline and scope
• Success criteria

Present to user:

• Confirm you've reviewed their plan
• Show the list of hypotheses you identified
• Ask if this matches their understanding

Example:

"I've reviewed your analysis plan for the 15% decline in 7-Day Publish Rate. I see 8 hypotheses organized into 3 tiers:

Tier 1 (High Priority):

- H1: Account Setup Completion Rate Dropped

- H2: First Prompt Submission Rate Declined

- H3: User Mix Shifted to Lower-Quality Signups

Tier 2 (Medium Priority):

- H4: Project Start Rate Decreased

- H5: Preview-to-Publish Conversion Dropped

- H6: Time-to-First-Prompt Increased Beyond 7 Days

Tier 3 (Lower Priority):

- H7: Help Feature Usage Changed with Different Outcomes

- H8: External/Seasonal Factor

Does this match your plan?"

Step 2: Hypothesis Triage Phase

Objective: Quickly map out which hypotheses were analyzed.

Approach:

• Ask user to identify analyzed vs not-analyzed hypotheses
• For not-analyzed items, get brief reason (optional but helpful)
• Keep this phase quick - just a checklist\
• Use the askquestion tool to facilitate input from the user

Example questions:

• "Which of these hypotheses did you actually analyze? You can just list the IDs like 'H1, H2, H3' or tell me which ones you skipped."
• "For the ones you didn't analyze, was it because earlier findings explained everything, or another reason?"

What to capture:

Analyzed: [List of hypothesis IDs]
Not analyzed: [List of IDs with optional brief reason]

Don't: Ask for detailed findings yet. Just get the map of what was done.

Step 3: Validation Status Phase

Objective: Understand which analyzed hypotheses were proven vs disproven.

For each analyzed hypothesis:

• Ask for validation result: Validated | Invalidated | Partially Validated
• Request confidence level: High | Medium | Low
• Get rough impact estimate if validated (e.g., "contributes X% to the decline")
• Must use the askquestion tool to facilitate input from the user on these 2 items (validation, confidence) and free text for the impact

Example questions:

• "For H1 (Account Setup Completion Rate Dropped), did your analysis validate or invalidate this hypothesis?"
• "How confident are you in this finding - high, medium, or low?"
• "If validated: Can you estimate how much this contributed to the overall problem? For example, if the overall decline was 15%, did this hypothesis explain 5%, 10%, or some other amount?"

Create a summary table:

| Hypothesis | Status | Result | Confidence | Impact |
|-----------|--------|--------|------------|--------|
| H1 | Analyzed | Validated | High | 8% of 15% |
| H2 | Analyzed | Invalidated | High | 0% |

Present this table to the user for confirmation before proceeding.

Step 4: Detailed Findings Collection

Read the conversation guide for comprehensive patterns: references/conversation_guide.md

Read the writing guide for quality standards: references/writing_guide.md

Core approach:

• Work through hypotheses ONE AT A TIME
• Gather information using SEBI framework: Statement, Evidence, Business Impact, Insight
• Focus on communication-ready content, not raw data dumps
• Adapt questions based on whether hypothesis was validated or invalidated

For VALIDATED Hypotheses

Information to gather:

1. Core finding (2-3 sentences): What actually happened?
2. Supporting evidence (3-5 data points): What proves it?
3. Impact quantification: How much did this contribute?
4. Business implication: Why does this matter?

Example conversation flow:

Claude: "Let's dive into H1 - Account Setup Completion Rate Dropped. You validated this hypothesis. Can you summarize what you found in 2-3 sentences that could be shared with leadership?"

User: [Provides summary]

Claude: "Great. What are the 3-4 most important data points that support this finding?"

User: [Provides data points]