SaaS Research — Idea Discovery & Validation

Executes docs/BOOTSTRAP.md Phase A (A1 idea search → A2 market research → scorecard for the A3 decision). Distilled from references/playbook.md (5 practitioner methodologies + 2025–2026 web research) — read it for depth; this file is the executable process.

The core loop (everything below is a variation of it)

Find where demand already exists but the solution is weak → narrow to a niche you can win → confirm people will pay → build the smallest thing.

Iron rules

1. Do not invent a novel idea in an unproven market. Competition = validation. The edge is positioning, distribution, and speed — not novelty. 0–1 competitors is a red flag, not a blue ocean.
2. AI reasons; metrics come from real tools. Never present model-estimated volumes/CPC/KD as fact — pull from an SEO MCP (DataForSEO / Semrush / Ahrefs / GSC) when configured, else free tools (Google Trends, Keyword Planner) via browsing, else mark [Unverified].
3. Never invent quotes. Pain evidence is verbatim + permalink, or it doesn't count.
4. The reusable opportunity filter: KD < 20–30 + CPC > $5–10 (buyer money) + commercial/pain intent + a weak SERP (page-1 with DA/DR<20, thin/stale content, or Reddit/Quora threads ranking = nobody owns it).
5. Distribution is the #1 silent killer — an idea without a reachable channel scores itself out, no matter how painful the problem.

Process

Step 1 — Pick hunting grounds (with the user)

Evergreen market (Health/Wealth/Relationships/B2B ops) or a vertical the founder knows; list the SaaS they already pay for (best seed per Gawley); default B2B, web-first, $3k–$50k MRR scale.

Step 2 — Run the four lenses in parallel (an idea surfacing in ≥2 lenses is a strong signal)

Lens A — SEO economics (most reproducible): keyword-gap vs 3–4 incumbents; hunt 3+ word "[tool] for [niche]" long-tails; apply the filter from rule 4; ALWAYS read the SERP for weakness (low-authority domains, stale content, forum threads at positions 4–10; bail if the whole top-5 has 50+ referring domains). Intent modifiers that signal ideas: alternative, migrate from, too expensive, is there a tool that, spreadsheet/manually, [software] for [profession], template/calculator/generator/tracker.

Lens B — Trends & waves: Google Trends → Rising/Breakout, Topics over terms, require a sustained 6–12-month uptrend (not a spike). Ask "why now?" — platform shifts (MCP servers, ChatGPT Apps SDK, voice agents), marketplace economics (Shopify 100%-of-first-$1M), and regulation deadlines (EU Accessibility Act, e-invoicing/ViDA, EU AI Act, US state privacy) are recurring-revenue wedges. Moat test — need ≥2 of: proprietary data, embedded workflow, regulatory lock-in, distribution.

Lens C — Community & marketplace pain mining: the universal tells are "is there a tool that…" / "I wish there was…" / "someone should build…" / "how do I automate…".

• Reddit via Google: site:reddit.com ("is there a tool" OR "i wish there was" OR "someone should build" OR "how do I automate") [niche] after:<last-18mo>
• Negative reviews (1–3★ = pre-validated demand): G2 "Lowest Rated" dislike fields, Capterra cons, Shopify ?ratings[]=1, Apple RSS JSON feed. 3–4★ "love it but missing X" maps the wedge.
• Marketplace gaps: high installs + <4.0★ + stale updates (Chrome MV3 die-off, WordPress plugin API, Zapier/Make missing-connector probing, HubSpot Ideas board).
• Freelance/job boards: people repeatedly PAYING for the same rule-based manual task = proven WTP.
• One complaint is noise; the same complaint across 10+ posts is a signal.

Lens D — Copy & reposition (lowest risk): proven product (IndieHackers $3k–$50k MRR, Acquire.com listings, top-grossing charts, a SaaS the founder pays for) + ONE differentiation move: niche-down · light/cheaper · local/language · compliance · combine 2–3 products · reverse-engineer the tech · productize a service.

Prompt library (grounded extraction, clustering, scoring, discovery queries): references/playbook.md Part 4 — use verbatim.

Step 3 — Shortlist & score (the A2 deliverable)

Weighted scorecard per idea (1–5 × weight, max 90): Pain intensity ×3 · WTP ×3 · Demand evidence ×2 · Competition health ×2 (2–10 competitors with broken execution = healthy) · Solo-build feasibility ×2 (MVP ≤4 weeks) · Distribution ×3 · Differentiation ×2 · Stickiness ×1. ≥70 build · 54–69 validate harder · <54 kill. Hard vetoes (any one = bail): pain=1 · distribution=1 · 0–1 competitors.

Step 4 — Outputs (contract with BOOTSTRAP)

• research/ideas.md — 5 ideas, each: one-liner · niche/ICP · the demand evidence (keyword numbers with source, pain quotes with permalinks, competitor revenue signals) · which lenses surfaced it · proposed wedge/positioning.
• research/market-<idea>.md ×5 — competitors + pricing norms, channel analysis, why-now, realistic MRR comps (median profitable micro-SaaS ≈ $4.2k MRR; ~70% never break $1k — set expectations honestly), scorecard row rationale.
• research/scorecard.md — the table, sorted, with vetoes flagged and a recommendation. The founder decides (gate G1); the skill recommends, never decides.

Step 5 — Pre-build validation assets (on request, after G1)

Fake-door landing page brief (one ICP, visible price, one CTA; pass bar 5–10% opt-in, <2% kill) · cold-outreach script (20 DMs; pass 10+ signups AND 3+ "I'd pay") · Mom-Test interview guide (past behavior only) · pre-sell framing (3+ paid commitments = build; waitlist signups don't count; exception: Starter Story's 100-waitlist bar for consumer). Set the pass bar BEFORE seeing data.

Tooling

WebSearch/WebFetch always. SEO MCPs when configured (DataForSEO cheapest, Semrush per the playbook's discovery prompt, GSC free on own data); Apify MCP for Reddit/review scraping; Google Trends free. Degrade gracefully — missing tool ⇒ narrower claim + [Unverified] marker, never a made-up number. Full tooling cheat-sheet: references/playbook.md Appendix.