Do competitor and market research with AI
Skills used in this guide (4)
上游市场研究方法论，自上而下与自下而上双算 TAM/SAM/SOM、调研样本量与细分评分。
npx skills add alirezarezvani/claude-skills -g构建证据可追溯的市场研究报告，以及基于假设的细分或预测情景。用于市场定义、行业与客户证据、竞争格局、TAM/SAM/SOM 对账、预测敏感性，以及可审计的报告骨架。
npx skills add k-dense-ai/claude-scientific-skills -g- npx skills add affaan-m/ECC -g
读研究大纲，为每个条目派独立 agent 深度研究。禁用 task output。
npx skills add Weizhena/Deep-Research-skills -g
The value of research is the source, not the length of the conclusion. The discipline: every key number carries a source link; conclusions AI can't source are guesses — keep them out of formal material.
Who this is for / not for
For:
- Quickly map a market / a class of competitors
- Need a cited brief to support a decision
- Willing to verify key numbers
Not for:
- Legal-grade due diligence (AI research can't replace it)
- Domains where data is paid / not public
- Want a pretty conclusion, don't care about basis
Before you start
- Research question: what to answer (size? competitor diff? barrier to entry?).
- Known info: your internal shares, data.
- Trusted source list: prefer official, broker, industry reports; avoid promo fluff.
Steps
1. Install the skill
market-research is a cited market-research skill. Copy its folder into your client's skills directory (same as any Agent Skill):
| Client | Directory |
|---|---|
| Claude Code | ~/.claude/skills/ |
| Codex | ~/.codex/skills/ |
| Cursor | ~/.cursor/skills/ |
| Trae | ~/.trae/skills/ |
| CodeBuddy | ~/.codebuddy/skills/ |
| Tongyi Lingma | ~/.lingma/skills/ |
2. Break the question into retrievable sub-questions
Research the domestic AI coding-assistant market, brief:
1. main players (5 domestic, 5 global) and positioning diff
2. estimate MAU tier from public data (cite source)
3. compare pain points across 3 user types
Every key number must carry a source link; unsourced marked "unverified", don't invent.3. Cross-check multiple sources
Use research-deep for multi-source deep retrieval; key numbers confirmed by ≥2 independent sources; flag big gaps.
4. Land on a decision
Conclusion must return to "so what should we do", not stop at description.
Pitfalls
| Pit | Symptom | Fix |
|---|---|---|
| No source | Pretty but untraceable | Force source link per key number |
| Single source | One voice as fact | ≥2 sources for key numbers |
| Mixed口径 | Forecast as actual | Distinguish actual / forecast / target |
| Promo as data | Numbers from fluff | Prefer official + third-party reports |
| Describe not conclude | Read but no action | Conclusion returns to "what to do" |
FAQ
Use AI numbers directly?
Not as fact. Every key number must carry a source; unsourced marked "unverified".
Formal due diligence?
AI research is screening and mapping; formal DD still needs experts + paid data.
Competitor compare without errors?
Use competitive-platform-analysis for platform compare, cross-check official sources.
How long?
First brief ~30 min; rigorous cross-check takes longer but worth it.
Track competitors on schedule?
Yes — make the research task run on a schedule, produce cited change briefs regularly.
Related guides
Continue across these scenes:
- Turn raw data into a report you can ship, with AI
- Write cold emails and outreach with AI
- Analyze papers and literature with AI
相关源码
- alirezarezvani/claude-skills · market-research source
- affaan-m/ECC · competitive-platform-analysis source
- Agent Skill open format spec
一句话结论
结论：调研的价值在出处不在长度——每条关键数字都带来源，给不出处的当猜测别写进正式材料。