Competitor Research

Competitor research

Run a 13-dimension dossier on a B2B SaaS competitor. Output ships with explicit confidence levels, inline source citations, a consolidated Sources & data quality table, and a data-gaps section with follow-up actions. Knowledge type: competitor-intel (per .claude/rules/ontology.md); maturity: emergent → validated after client review.

Research substrate

Default substrate: Exa (per .claude/rules/exa-protocol.md, auto-loaded). Primary tools: find_similar_links_exa (structural competitor discovery from a client URL), web_search_exa (news, voice, gap discovery), web_fetch_exa (clean comparison-page extraction). Crawl-cost discipline (.claude/rules/crawl-cost-discipline.md): when a dimension needs a competitor's page inventory (Content, SEO/AEO), enumerate pages first with free mcp__spider__spider_links (local, no credits), triage to product / pricing / positioning / customer pages, then spend web_fetch_exa / Firecrawl only on kept pages — this kills the "fallback to Apify if credits exhausted" failure on large sites. Migration window: prefer the plugin namespace mcp__plugin_exa_exa__* once installed; legacy mcp__exa__* still mounted as fallback. Worked examples + tool catalog: .claude/skills/meta-skills/exa/.

When to run

Invoke when the user asks for: competitor analysis for X, battlecard research for X, competitive landscape for [market], compare X vs Y, what's [competitor] doing?. Do NOT invoke for: company qualification (use /company-context), product messaging only (use /product-messaging), researching the user's own company (use /company-context), or single-feature questions (answer directly).

Three run modes — pick by cadence and depth needed:

• First run (~90 min) — /competitor-research [competitor]. New competitor, no existing dossier. Phases 1–3 (or 1–4 for aggregate).
• Quick scan (~30 min, weekly) — /competitor-research --quick [competitor] or /loop 1w. Refreshes fast-moving data only (news, Clay, G2, internal sources). Updates "Recent changes" header — sections without changes stay untouched.
• Deep refresh (~90 min, monthly) — /competitor-research --refresh [competitor] or /loop 1M. Full 13-dimension cycle with TrustPilot monitor and Phase 4 aggregate (if 2+ competitors refreshed this cycle).

Full cadence + refresh discipline → the premium reference. Visual phase map → the premium reference.

The Iron Law: no data point without source. Every claim cites a URL + access date or is marked [Not available]. Estimates need explicit confidence + reasoning. "Fast + wrong = useless."

Inputs

Required:

• competitor name — exact company/product name (e.g., Linear, Lovable).
• website URL — competitor's site (e.g., linear.app).

Recommended (improve quality):

• client context — current positioning sharpens the comparative angle.
• specific questions — focus areas (e.g., "deep dive on their pricing model and outbound motion").
• research mode — single deep dive (default) or comparison matrix (3–6 competitors).

If competitor name is ambiguous (e.g., "Cursor", "Base", "Linear"), confirm with the user before starting. If website URL is missing, ask — don't guess.

Steps

1. Confirm scope and disambiguate → the premium reference. Verify name + URL, select run mode, lock disambiguation.
2. Pre-flight optional MCPs → check Clay, Granola, Google Drive, CRM, Notion availability per the premium reference ("Optional internal intel sources"). For each available, queue the corresponding queries; for each missing, skip silently.
3. Research 13 dimensions → the premium reference. Order: Company → Product → ICP → Pricing → Reviews → Content → Launches → SEO/AEO → Technographics → Openings → GTM → LinkedIn/Social → Paid advertising. Each dimension has tool fallbacks (Ahrefs → Serper → manual; Apify → Firecrawl → manual). Frameworks + scoring → the premium reference ("Core frameworks").
4. TrustPilot customer monitor (deep refresh only, customer-facing competitors only) → run inside Step 2.5 per the Phase 2 file. Score top 10–15 customers 🟢/🟡/🔴.
5. Synthesize and assign confidence → the premium reference. Assign [VERIFIED]/[INFERRED]/[ESTIMATED]/[UNAVAILABLE] per claim, write 2–3-paragraph executive summary, document data gaps with follow-up actions.
6. Aggregate analysis (only if 2+ competitors researched for same client) → the premium reference. Build threat matrix, feature parity table, credibility audit, 3–5 strategic recommendations. Save as MMYY-aggregate-insights.md in client's competitors folder.
7. Apply attribution standard — inline (Source: X) per claim (not the verbose [VERIFIED: url, date] block); consolidate full URLs + access dates + confidence in the Sources & data quality table at the end of the document. Quality threshold: ≥50% Verified, ≤20% Estimated.
8. Write to client folder per output template → the premium reference ("Inline canonical, v2.7" — 13 dimensions, TrustPilot sub-section, Recent changes header, Sources & data quality table). For matrix mode → the premium reference ("Compact alternative"). Search query patterns per dimension → the premium reference. Common source URL patterns → the premium reference.
9. Self-evaluate against quality gates → the premium reference. Run completeness, evidence-quality, and guardrail checks before declaring "done".
10. Push to Notion (Competitor Research Database) and Google Docs (client_folder/context/competitor-research/) per the push targets in frontmatter. For refresh mode, update the existing Notion page rather than creating a duplicate.
11. Offer iteration prompts post-delivery → the premium reference. If user signals approval ("great research", quick approval), offer to save the output as a reference example under the premium reference.

What good looks like

Evaluations (binary pass/fail before declaring "done")

• ≥3 sources per major claim (revenue, funding, team size, pricing).
• ≥50% [VERIFIED] confidence; ≤20% [ESTIMATED].
• Threat level set on every competitor: PRIMARY / ENTERPRISE TIER / DIRECT ICP / STEALTH WATCH / LOW / DEFUNCT.
• Inline (Source: X) on every key data point (no verbose [VERIFIED: url, date] blocks inline).
• Sources & data quality table at end-of-document with URLs + access dates + confidence per dimension.
• Data-gaps section non-empty if any dimension is incomplete or [UNAVAILABLE].
• All 13 dimensions present (or marked [Not available] / [Requires premium tool] explicitly).
• Recent-changes header present in --quick and --refresh modes; "Last refreshed" date updated.
• Output title is # Competitor research: [Name] exactly — no aliases like "competitive intelligence analysis" or "Competitive Intelligence Report".
• Premium-tool limitations noted explicitly (Ahrefs absent → Serper or manual, BuiltWith absent → "Not available", JS-walled ad libraries → manual check noted).

Integration with other skills

Recommended chains:

• Comprehensive competitive analysis: company-context → competitor-research → positioning.
• Decisions-to-make deck: competitor-research → positioning --scenarios — append the positioning options + canvas as the deck's "next steps / decisions to make" closing section.
• Battlecard creation: competitor-research → sales-enablement.
• Full PMM stack: icp-behavioural + competitor-research → positioning → product-messaging → sales-enablement.

Scheduling (recurring runs)

/loop 1w /competitor-research --quick [competitor-name] # weekly quick scan, set-and-forget
/loop 1M /competitor-research --refresh [competitor-name] # monthly deep refresh, review-and-approve

/schedule create --name "competitor-[name]-weekly" --cron "0 9 * * 1" --prompt "/competitor-research --quick [name]"
/schedule create --name "competitor-[name]-monthly" --cron "0 9 1 * *" --prompt "/competitor-research --refresh [name]"

For multiple competitors in parallel: /competitor-research --refresh --all spawns one subagent per competitor in the premium reference.

Pre-slim original

Pre-slim SKILL.md (1609 lines, v2.7) archived at .claude/skills/_archive/competitor-research/SKILL-pre-slim-20260429.md. See the premium reference for the v2.8 changelog entry documenting the slim.

Sourced patterns — person-level competitor profiling

Standard /competitor-research profiles the company (11 dimensions covering positioning, pricing, features, content, GTM, team, funding, etc.). For deeper buyer-side intelligence, add a person-level layer that profiles the named decision-maker / champion / founder behind the competitor:

• Founder / CEO behavioral profile. Public-content cadence (LinkedIn, X, podcast appearances), preferred narratives, hiring-page tells (what roles they're scaling), customer-conversation patterns (testimonials they highlight, objections they pre-empt). Read for positioning intent — founders telegraph strategy in public.
• Champion persona at the buying account. Who actually drives the buy at their target accounts? Title patterns, tenure patterns, pain-point language (lifted from their case studies). This is the ICP-buyer side of the equation, not just the ICP-company side.
• Departure / hiring signals. Recent senior departures at the competitor = signal of internal tension; recent senior hires = signal of investment lane. Both inform competitive timing.
• Social-graph cluster. Who are they engaging with on LinkedIn? Who endorses them publicly? Builds a map of allied / adjacent companies that may be partnership or co-marketing candidates (cross-references with /co-marketing).

Use when the standard 11-dimension company profile isn't enough — typically for direct-competitor showdowns or accounts where the champion person is the deal-maker (mid-market and below).