deep-research

Deep Research Methodology

Systematic approach to web-based research for architecture and product decisions. Use this skill whenever generating market research, competitor analysis, pitch decks, cost estimates, or technology evaluations.

---

When to Use This Skill

• Generating competitor analysis or market landscape
• Estimating market size (TAM/SAM/SOM)
• Validating technology choices against current ecosystem
• Producing investor-facing materials with market data
• Updating cost estimates with current pricing

---

Three-Phase Research Process

Phase 1: Discover (Broad Search)

Cast a wide net to identify the landscape:

1. Competitor search: Use WebSearch for:

   • "{product category}" alternatives
   • "{product category}" competitors 2025
   • "best {product category}" software
   • "{product category}" market leaders

2. Market sizing search: Use WebSearch for:

   • "{product category}" market size 2025
   • "{product category}" TAM SAM
   • "{industry}" market report
   • "{product category}" growth rate forecast

3. Funding/trends search: Use WebSearch for:

   • "{product category}" startup funding 2024 2025
   • "{product category}" trends
   • "{competitor name}" funding round

4. Collect 10-15 relevant URLs from search results for Phase 2.

Phase 2: Verify (Targeted Validation)

Drill into specific sources to extract and verify data:

1. For each competitor found — Use WebFetch to:

   • Visit their homepage → extract value proposition, target audience
   • Visit pricing page → extract pricing tiers, free tier limits
   • Visit about page → extract founding date, team size, funding

2. For market size claims — Cross-reference with 2+ sources:

   • Industry reports (Gartner, Statista, Grand View Research)
   • Press articles citing market data
   • If sources conflict, note the range and which source is more authoritative

3. For funding data — Check:

   • Crunchbase profiles
   • TechCrunch or press release announcements
   • LinkedIn company pages for employee count estimates

4. For technology claims — Verify:

   • GitHub star counts (WebFetch github.com/{repo})
   • npm download counts for packages
   • Stack Overflow survey data

Phase 3: Synthesize (Structured Output)

Organize findings into a clear, actionable document:

1. Confidence tagging — Mark every data point:

   • [Verified] — Confirmed via direct source (pricing page, press release, report)
   • [Estimated] — Based on partial data or extrapolation
   • [Inferred] — Based on training data, not web-verified

2. Source citation — For every verified claim:

   Market size: $4.2B in 2025 [Verified]
   Source: Grand View Research, "Collaboration Software Market Report 2025"
   URL: https://example.com/report
   

3. Data freshness — Note when data was published:

   Last verified: March 2025
   Source published: January 2025
   

4. Structured tables — Present competitor data as scannable tables:

   Competitor	Pricing	Funding	Key Feature	Weakness
   {name}	{tier}	{$Xm}	{feature}	{gap}

Phase 4: Advanced Analysis

Calculate quantified metrics beyond basic research:

1. Competitive Momentum Scoring (0-100 scale)

   • Funding growth trajectory: How much faster are they growing in capital? (+50% = high acceleration)
   • Hiring velocity: Engineers added per year relative to current team
   • Release frequency: How often do they ship? (Weekly = active, Monthly = maintenance mode)
   • Formula: (Funding growth × 0.35) + (Hiring velocity × 0.35) + (Release frequency × 0.3)
   • Result: Identifies threats by acceleration, not just current size

2. Feature Parity Scoring (0-100 scale)

   • Count features in each category vs competitors
   • Calculate: (Our Features / Avg Competitor Features) × 100
   • <50% = Playing catch-up
   • 50-80% = Competitive, can win on execution
   • 80%+ = Feature parity, differentiate elsewhere

3. Market Sizing with Confidence Intervals

   • Weight multiple sources by credibility: Gartner (0.4) > Statista (0.3) > Blog (0.15) > Extrapolation (0.15)
   • Calculate weighted average, not just averaging estimates
   • Provide range (80% confidence): If sources suggest $3.8B-$4.8B, report the range
   • Sensitivity: "If market grows 8% vs 12% vs 18% CAGR, our 3-year opportunity is..."
   • Output: "$4.2B TAM ($3.8B-$4.8B range)" not "$4.2B TAM"

---

Research Quality Rules

1. ALWAYS use WebSearch before making market size, competitor, or pricing claims
2. ALWAYS cite sources with URLs for every verified data point
3. NEVER present estimated data as verified fact — use confidence tags
4. If WebSearch is unavailable, explicitly state:

   "Note: Based on training data as of [date]. Live web search was not available. Data should be independently verified before use in investor materials."

5. Prefer recent sources — prioritize data from the last 12 months
6. Cross-reference critical claims with 2+ independent sources
7. Use WebFetch on competitor websites directly — don't rely solely on third-party descriptions
8. Note limitations — if a competitor's pricing page requires login, note "pricing not publicly available"

---

Output Structure Template

# [Research Topic] — Deep Research Report

**Project**: [Project name from SDL]
**Researched**: [Today's date]
**Web search**: [Used / Not available]

## Executive Summary
[3-5 key findings]

## Competitor Landscape
[Table of competitors with verified data]

## Market Sizing
[TAM/SAM/SOM with sources]

## Feature Comparison Matrix
[Our product vs competitors across key dimensions]

## Opportunity Gaps
[Underserved areas competitors miss]

## Technology Landscape
[Relevant tech trends and adoption]

## Sources
[Numbered list of all URLs referenced]

---

Integration with Other Commands

This skill is automatically loaded by:

• /architect:deep-research — Full standalone research report
• /architect:cost-estimate — For verifying current cloud/service pricing
• Referenced by Ideation Playbook actions (Market Research, Pitch Deck)

Other commands can reference this skill when they need web-verified data:

### Step N: Load Skills
Load **deep-research** skill for web research methodology