research-analysis

Research Analysis

Overview

This skill provides McKinsey-style consulting analysis capabilities, including structured thinking frameworks, quantitative modeling methods, competitor analysis templates, user research methodologies, and tool integration guides. It transforms complex business problems into conclusion-first, data-supported, actionable recommendations.

Core Principles

All analysis must follow three rules:

1. Conclusion first - Lead with the answer, then support with data (pyramid principle)
2. Data supported - Every claim backed by evidence with cited sources
3. Actionable - End with specific, prioritized recommendations with timelines

Analysis Workflow

Step 1: Problem Definition

Clarify the core question before diving into analysis:

• Restate the user's question in SCQA format (Situation, Complication, Question, Answer)
• Identify the decision-maker and what decision they need to make
• Define the scope: what is in-scope vs. out-of-scope
• Determine the output format: executive summary, full report, presentation, or comparison table

Step 2: Framework Selection

Load references/analysis-frameworks.md and select the appropriate framework(s):

Multiple frameworks can be combined for complex problems.

Step 3: Data Collection and Analysis

Based on the analysis type, load the relevant reference files:

• Competitor analysis -> Load references/competitor-analysis.md

  • Use the 6-dimension framework (product, business model, market, company, operations, reputation)
  • Query enterprise data via Tianyancha MCP (see references/tool-usage.md)
  • Query app market data via Qimai (see references/tool-usage.md)

• User research -> Load references/user-research.md

  • Select research method based on goal (qualitative vs. quantitative)
  • Build user personas using the 3-dimension framework (demographic, behavioral, psychological)
  • Map user journeys with emotion curves and opportunity points
  • Segment users using RFM or lifecycle models

• Quantitative modeling -> Load references/data-modeling.md

  • Build metric systems (North Star metric -> core metrics -> process metrics)
  • Estimate market size using TAM/SAM/SOM or Fermi estimation
  • Calculate unit economics (LTV, CAC, payback period)
  • Analyze funnels, cohorts, and ROI
  • Perform sensitivity analysis on key variables

Step 4: Data Visualization

Load references/tool-usage.md (VisActor section) and generate charts:

Chart rules:

• Title states the conclusion, not the data description
• Annotate data source and date below each chart
• Use professional color palette; Chinese stock convention: red = up, green = down
• Generate as standalone HTML with VisActor CDN for browser rendering

Step 5: Report Generation

Structure the final output as:

1. Executive Summary (3-5 key findings + top 3 recommendations)
2. Background & Methodology
3. Analysis (framework-driven, data-supported)
4. Key Findings (with visualizations)
5. Action Recommendations (prioritized table: P0/P1/P2 with timeline)
6. Data Sources & Limitations

Tool Integration

Tianyancha (Enterprise Data)

When enterprise information is needed (competitor background, financing, shareholders, IP, risk):

1. Check if tyc-mcp connector is connected (connector name: tyc-mcp Tianyancha)
2. If connected: call mcp__tyc-mcp__* tools to query enterprise data
3. If not connected: inform user to enable it in Connector Management, or use WebSearch to search Tianyancha public pages as fallback

Qimai (App Market Data)

When app market data is needed (downloads, rankings, reviews, ASO):

1. Use WebSearch to find Qimai data pages for the target app
2. Use WebFetch to extract data from public Qimai pages
3. Alternatively, accept user-provided Qimai export files (Excel/CSV)

VisActor (Charts)

Generate standalone HTML files with VisActor CDN for data visualization:

• VChart CDN: https://unpkg.com/@visactor/vchart/build/index.min.js
• VTable CDN: https://unpkg.com/@visactor/vtable/build/index.min.js
• Full code templates available in references/tool-usage.md

Information Quality Standards

1. Multi-source verification: Key data points verified by at least 2 independent sources
2. Source attribution: Every data point annotated with source and date
3. Fact vs. opinion: Objective data separated from analyst opinions
4. Timeliness: Prioritize data from the last 6 months
5. Confidence labeling: Mark data confidence as high/medium/low

Common Data Sources

Bundled Resources

Changelog

v1.0.0

• Initial release
• 10 consulting frameworks (MECE, SCQA, Pyramid, SWOT, Porter's Five Forces, 3C, 4P, BMC, Hypothesis-Driven, Kano)
• 6-dimension competitor analysis methodology
• User research toolkit (personas, journey maps, RFM, lifecycle)
• Quantitative modeling (TAM/SAM/SOM, unit economics, funnel, cohort, ROI)
• Tool integration guides (Tianyancha, Qimai, VisActor)