| name | market-research |
| description | Compile structured market/competitive intelligence from web sources into actionable reference documents. Use for landscape scans, vendor comparisons, pricing research, industry mapping, and reconnaissance tasks where the deliverable is a synthesized document, not just an answer.
|
| triggers | ["research X market/industry/landscape","find who does X and compare them","compile a list of vendors/services/providers for X","what are the options for X","competitive intelligence or pricing research requests"] |
Market Research & Intelligence Compilation
Workflow
1. Scope the Research
- Identify the key dimensions the user cares about (pricing, features, coverage, payment terms, etc.)
- Note any specific filters (geography, tier, size, etc.)
2. Multi-Query Search Strategy
Execute 3–6 independent searches in parallel, varying the angle:
- Direct name searches ("X listing agents services")
- Specific feature searches ("X accept payment in Y")
- Comparison/ranking searches ("best X 2024 2025 comparison")
- Cost/fee searches ("X pricing fees cost breakdown")
Batch independent searches in a single turn. Each search returns snippets — read them carefully for cross-references and specific data points.
3. Deep Extraction (when available)
- Use
web_extract on promising URLs for full-page content
- Fallback: When
web_extract fails (rate limits, bot blocking), search snippets often contain enough detail — cross-reference across multiple snippet sources to fill gaps
- Pitfall: Don't retry
web_extract more than twice on the same domain in one turn — it's likely rate-limited. Move to alternative search queries instead.
4. Cross-Reference & Validate
- Look for the same data point appearing in 2+ independent sources
- Flag single-source claims as "reported" or "estimated"
- Note when information is clearly outdated (check dates in snippets)
5. Compile Deliverable
Structure the output as a markdown document with:
- Summary table(s) for quick scanning
- Tiered/grouped organization (by size, cost, capability, etc.)
- Explicit confidence markers (confirmed vs. estimated vs. rumored)
- Actionable takeaway section at the end
Save to /tmp/ with a descriptive filename. The user typically wants:
- A document they can reference/share, not just a chat answer
- Clear structure that lets them find specific items fast
- Honest gaps — don't fill unknowns with plausible-sounding guesses
Pitfalls
- Don't stop at the first search — surface breadth matters for market research
- Don't fabricate pricing or contact details you didn't find
- Don't omit negative findings ("X does NOT charge fees") — they're often the most useful data points
- When web_extract fails, compile from snippets — the research doesn't stop
