← ClaudeAtlas

res-competitor-researchlisted

Competitor research — systematic profiling of rivals' product, pricing, positioning, GTM, and moat with evidence-graded comparison tables. Use when user says "competitor analysis", "competitive research", or .
meshcode-ai/skills-research · ★ 1 · Code & Development · score 63
Install: claude install-skill meshcode-ai/skills-research
# Competitor Research ## Fixed scope - Limit to 3–5 direct competitors (same segment) + 1–2 indirect competitors (substitutes). Everything else goes in an appendix. - Fix comparison axes to **your own criteria** — comparing on the axes competitors publish locks you into their frame. ## Collection priority (confidence hierarchy) 1. Competitor's own: pricing page, product docs, **job postings (a mirror of strategy)**, changelog, filings & investor materials 2. Customer voice: 100–200 reviews (G2, app stores, Reddit) — not the star average, but the **3 most-repeated complaints** are the real asset 3. Third-party: analyst reports, trade media, partnership pages. Always note the discounted price alongside list price. ## Comparison criteria - **Positioning**: "what it does not do" (features excluded, segments excluded) is the substance of positioning. - **Moat verdict**: which of brand / switching cost / data / network / supply cost it leans on + durability signals (key personnel departures, share of new entrants). - **Price comparison**: convert to a common unit (per user, per usage, per transaction). If conversion is impossible, state "not comparable". - Every figure needs a source URL + verification date. Never promote a competitor's marketing copy into fact. ## Output ``` # Competitor Research — {category} ({date}) ## Comparison table (axis × competitor, source & date in every cell) ## Relative position (2 strengths · 2 weaknesses · 1 unoccupied terrain) ## Top 3 competitor