| name | market-research |
| description | Conduct market research, competitive analysis, investor due diligence, and industry intelligence with source attribution and decision-oriented summaries. Always activate when the user wants market sizing, competitor comparisons, fund research, technology scans, TAM/SAM/SOM estimates, or any research that informs a business decision. Also activate when the user says "research X", "what do you know about Y company", "compare these competitors", "who should I talk to about Z", or "help me think through this market" — even without the word "research". |
Market Research
Produce research that supports decisions, not research theater. Every deliverable should make a specific decision easier for a specific person.
Workflow
When this skill activates:
- Ask the decision question first — before searching anything, ask: "What decision will this research support?" The answer determines mode, depth, and output format. One question; don't stall.
- Choose the research mode from the taxonomy below. Modes can combine.
- Build a search plan — identify what to look for and where before searching. See Signal Sources below.
- Search in passes — broad first, then targeted. Use web search throughout; don't rely on training knowledge for current market facts, funding rounds, or pricing.
- Separate fact from inference — label everything. Never present an estimate as a fact.
- Synthesize toward the decision — findings that don't bear on the decision get cut.
- Run the quality gate before delivering.
When to ship vs. do another pass: if the next pass would change the recommendation, keep going. If it would only add more supporting evidence, ship.
Research Standards
- Source every important claim — include URL and access date where possible
- Flag stale data explicitly — anything older than 18 months in a fast-moving market warrants a caveat
- Separate fact, inference, and estimate — use labels:
[fact], [inferred], [estimate]
- Include the bear case — find the strongest argument against your emerging thesis and address it directly
- Translate to a decision — the last thing you write should be: "Based on this, I would / would not [action], because..."
Research Modes
Investor / Fund Diligence
Purpose: Determine whether this fund is worth pursuing and how to approach them.
What to find:
- Fund size, stage focus, and typical check size (Crunchbase, fund website, PitchBook if available)
- Relevant portfolio companies — especially direct overlaps and adjacent bets
- General partner backgrounds and stated thesis (Twitter/X, Substack, podcast appearances)
- Recent investment activity — what they've funded in the last 6–12 months signals current priorities
- Known anti-portfolio signals — sectors or models they've publicly passed on
Fit assessment: after gathering, score the fit across three axes:
- Stage fit (are we at the right stage for their fund?)
- Thesis fit (does our category match their current focus?)
- Proof point fit (do we have what they typically need to write a check?)
Deliverable: a one-page dossier per fund with a clear fit / no-fit / maybe verdict and a tailored outreach angle.
Competitive Analysis
Purpose: Understand the competitive landscape well enough to position accurately and anticipate moves.
What to find:
- Product reality (not marketing copy) — find user reviews on G2, Capterra, Reddit, App Store, and Trustpilot for unfiltered product experience
- Pricing structure — look for pricing pages, teardowns, community discussions, and sales Reddit threads
- Distribution model — where do they get customers? Outbound, PLG, marketplace, partnership?
- Funding and investor history — Crunchbase, TechCrunch, their blog
- Team and hiring signals — LinkedIn headcount trends, active job postings reveal strategic priorities
- Technical signals — GitHub (open source activity, stars, issue velocity), StackOverflow tags, developer community
Positioning gaps: identify what every competitor is bad at (per real user reviews) that your target customers care about. That's your opening.
Deliverable: comparison matrix (you vs. each competitor on 6–8 dimensions that buyers actually care about) plus a narrative on where the gap is and whether it's defensible.
Market Sizing
Purpose: Establish a credible range for the addressable market — not a confident number, a reasoned range with explicit assumptions.
Approach — run both and reconcile:
Top-down:
- Find total industry revenue from analyst reports, public filings, or trade associations
- Apply segmentation percentages to isolate your addressable slice
- Express as a range (reports rarely agree — use the spread as your uncertainty band)
Bottom-up:
- Identify your ideal customer profile precisely
- Estimate the number of such customers (use LinkedIn company filters, industry databases, government data)
- Multiply by realistic ACV or annual spend
- Apply a realistic win rate and ramp timeline
Reconciliation: if top-down and bottom-up are more than 5× apart, you have an assumption problem — find it before presenting.
Label every number: [analyst estimate, 2024], [our calculation], [assumption: 15% win rate]
Deliverable: TAM / SAM / SOM in a table with a row per assumption, labeled by source and confidence.
Technology / Vendor Research
Purpose: Determine whether this technology or vendor is the right fit, and what the risks are.
What to find:
- Technical reality — find independent technical assessments, not vendor documentation. Look for engineering blog posts, conference talks, and community discussions
- Adoption signals — GitHub stars, npm/PyPI downloads, Stack Overflow activity, job posting volume using the technology
- Integration complexity — find migration stories and integration war stories on Hacker News, Reddit, and engineering blogs
- Lock-in risk — can you leave? What does egress cost? Is your data portable?
- Security and compliance posture — SOC 2, ISO 27001, GDPR, HIPAA — check their trust page and any known CVEs
- Vendor health — funding, headcount trend, customer references, churn signals
Build vs. buy signal: if integration complexity exceeds two sprints or lock-in risk is high, flag the build option explicitly.
Deliverable: a fit assessment with explicit go / no-go / evaluate-further verdict and the key risks if you proceed.
Thesis Pressure-Test
Purpose: Before committing (building, funding, entering), find the strongest argument against the thesis and determine whether it's fatal.
What to find:
- Who has tried this before and failed — and why (not just "bad execution")
- What structural forces would make this hard regardless of execution quality
- What would have to be true for this to work — and how likely each assumption is
- Who is well-positioned to copy this if it succeeds
- What the market looks like if the strongest competitor wins
Steelman the bear case: write the most compelling version of why this is a bad idea. Then address it directly. If you can't address it, that's the answer.
Deliverable: a two-column document — bear case and rebuttal — plus a summary verdict on whether the thesis holds.
Source Quality Hierarchy
Not all sources are equal. Apply appropriate skepticism:
| Tier | Sources | Trust Level |
|---|
| 1 — Primary | SEC filings, official company reports, government data, academic papers | High — verify the filing date and context |
| 2 — Analyst | Gartner, Forrester, IDC, CB Insights, Pitchbook | High for sizing, moderate for predictions |
| 3 — Quality press | FT, WSJ, Bloomberg, Reuters, The Information | High for facts, low for forecasts |
| 4 — Trade press | TechCrunch, VentureBeat, The Verge | Moderate — often based on press releases |
| 5 — Community | G2, Reddit, Hacker News, Twitter/X, Blind | Low reliability, high signal-to-noise on real sentiment |
| 6 — Company-owned | Press releases, blog posts, marketing copy | Always confirm independently; treat as advocacy |
Cross-tier rule: any important claim that rests only on Tier 4–6 sources needs a caveat or additional corroboration.
Indirect Signals Worth Searching
These are often more revealing than official sources:
Job postings — what a company is hiring for reveals strategic priorities 6–12 months before announcements. A sudden surge in sales engineers signals a go-to-market push. Hiring ML engineers at an infrastructure company signals a product pivot.
LinkedIn headcount — growth rate over the past 12 months signals runway and momentum. Decline signals trouble. Filter by function to see where they're investing.
GitHub — open source repos reveal technical architecture, activity level signals investment, issue tracker reveals product quality and responsiveness, star/fork velocity signals developer interest.
G2 / Capterra / App Store reviews — sorted by recency and lowest ratings. The complaints reveal real product weaknesses that positioning copy never will.
Patent filings — leading indicator of R&D direction. Search Google Patents by assignee.
Hacker News "Ask HN" and discussions — search site:news.ycombinator.com [company name] for unfiltered technical and founder community opinion.
Conference talks and podcast appearances — executives reveal strategy and thinking that press releases don't. Search YouTube and Spotify for founder and exec appearances in the past 12 months.
Output Format
Calibrate length to the decision's stakes. A quick competitive scan is different from pre-board due diligence.
Standard Report Structure
## [Research Subject]: [One-Line Verdict]
### Executive Summary
2–4 sentences. The decision, the strongest supporting evidence, the main risk.
No new information — only what's in the report below.
### Key Findings
Findings that bear on the decision. Labeled [fact], [inferred], or [estimate].
Source attributed per claim. Not exhaustive — only decision-relevant.
### [Mode-specific section]
For competitive analysis: comparison matrix
For market sizing: TAM/SAM/SOM table with assumptions
For investor diligence: fund fit assessment
For technology: build/buy/evaluate assessment
### Bear Case
The strongest argument against the thesis or recommendation.
Addressed directly, not dismissed.
### Recommendation
"Based on this research, I would [action] because [1–2 key reasons]."
If confidence is low: "I cannot recommend without [specific missing information]."
### Sources
Numbered list with URL, publication, access date, and tier (1–6).
Length Guidelines
| Scope | Length |
|---|
| Quick scan (single company, single question) | 400–700 words |
| Standard research memo (one mode, one decision) | 700–1400 words |
| Deep diligence (pre-investment, pre-entry) | 1500–3000 words, split into sections |
Quality Gate
Before delivering, confirm:
