docs: add GM-refined canonical strategy from copilot vision

docs/strategy/CANONICAL_STRATEGY.md

@@ -0,0 +1,372 @@
+# geolocal.io — Canonical Strategy
+
+> **Status:** Refined GM strategy (2026-07-18)  
+> **Source of truth for vision:** `docs/geolocal-copilot-conversation.md`  
+> **This document:** Condenses, stress-tests, and operationalizes that vision with current market evidence.  
+> **Supersedes:** Fractured narratives in older GTM/investor/engineering docs where they conflict.
+
+---
+
+## 1. One-line company definition
+
+**geolocal.io is the business-owned MCP infrastructure that makes local service businesses discoverable, interpretable, bookable, and measurable inside every AI assistant — the Shopify Storefront MCP equivalent for the long tail of local services.**
+
+Not a destination site.  
+Not a chatbot sold to SMBs.  
+Not a Yelp/Google clone.  
+Infrastructure that AI agents use, that businesses own, that partners and tourism boards distribute.
+
+---
+
+## 2. The problem (evidence-checked)
+
+### 2.1 Demand side: consumers already moved
+
+- BrightLocal’s 2026 Local Consumer Review Survey: **~45% of consumers use AI tools for local business recommendations** (ChatGPT is the leader among tools cited).  
+- BrightLocal and secondary coverage also reference a much lower prior-year figure (~6%); treat the directional leap as real, but do **not** overclaim a perfect YoY methodology match in investor materials without footnoting survey differences.  
+- X / practitioner discourse (2026): local discovery is collapsing into **1–3 named recommendations**, not a page of links — “if you’re not named, you’re invisible.”
+
+### 2.2 Supply side: AI is radically selective
+
+- SOCi 2026 Local Visibility Index (≈350k locations / 2,751 multi-location brands):  
+  - **ChatGPT recommended ~1.2% of brand locations**  
+  - Gemini ~11%, Perplexity ~7.4%  
+  - vs ~36% appearance in Google local 3-pack  
+- Locations recommended by ChatGPT skew high-trust (**~4.3★ average**).  
+- BrightLocal research: **Yelp is a frequent source** in AI local answers (~1/3 of searches in one study); listings/citations regained importance under LLMs.
+
+**Implication:** Traditional local SEO is necessary but insufficient. Structured, machine-actionable presence is becoming the bottleneck.
+
+### 2.3 Structural gap incumbents will not fill for SMBs
+
+| Player | What they are building | What they are NOT building |
+|--------|------------------------|----------------------------|
+| **Shopify** | Storefront MCP live on every store (`/api/mcp`); UCP with Google; Catalog MCP | MCP for non-Shopify local service businesses |
+| **Yelp** | Official open-source **Yelp MCP** over Fusion AI — *their* data for agents | Business-owned endpoints businesses control |
+| **Google** | Maps Grounding / Merchant MCP (alpha); AI in Maps/Search; UCP commerce | Neutral, multi-tenant “own your AI presence” for long-tail SMBs |
+| **Cal.com** | Full **booking MCP** (create/reschedule/cancel/availability) | Discovery, story, genre primitives, tourism aggregation |
+| **OpenAI / Anthropic / others** | Agent platforms + data licensing | Local service graph for the 98%+ long tail |
+
+**Thesis (validated):** Incumbents build MCP for *their* platforms and *their* data. The long-tail service economy (Bob’s Garage, local charters, salons, tourism boards) has no Shopify-equivalent on-ramp. That vacuum is geolocal.io.
+
+---
+
+## 3. North Star (from the copilot conversation — preserved)
+
+**Make AI give the most relevant and helpful local recommendation — and fulfill it — for local service businesses.**
+
+Helpfulness (not mere relevance) is the optimization target: clear services, specialization, pricing signals, availability, booking path, trust, and **repeatable success** so AI engines prefer the same path next time.
+
+---
+
+## 4. Four business models (do not collapse these)
+
+The copilot conversation defines **four stacked models**. Earlier repo docs mostly documented Model 1. That fracture is fixed here.
+
+| # | Model | Who pays / who adopts | What geolocal provides | Horizon |
+|---|--------|------------------------|------------------------|---------|
+| **1** | **Endpoint enablement** | Individual SMBs (self-service) | Hosted multi-tenant MCP + `/.well-known` pointer + Self-Service Portal (SSP) + diagnostics + reports | Day 0 |
+| **2** | **Intermediate enablement** | Tourism boards, chambers, visitor bureaus | Aggregating MCP for destinations/members; member onboarding; municipal ROI narratives | V1 GTM wedge |
+| **3** | **Industry trust layer** | Standards / partnerships / platforms | Quality primitives, decommissioning, genre norms — become a preferred discovery surface AI engines learn to trust | 12–24 mo |
+| **4** | **Service graph** | Data products / licensing | Network of related services, specialties, demand signals, competitive intelligence (“Citysearch 2.0, AI-native”) | After critical mass |
+
+**Rule:** Product decisions must state which model they serve. Do not build Model 4 features before Model 1–2 work.
+
+---
+
+## 5. Product definition (what we actually sell)
+
+### 5.1 What we sell
+
+**AI-readiness infrastructure for local services:**
+
+1. **Hosted multi-tenant MCP** — structured tools AI agents call (story, services, hours, booking path, related businesses, genre primitives).  
+2. **Dead-simple discovery pointer** — business drops a tiny well-known JSON / path on their site that points at geolocal (Shopify pattern: endpoint on *their* domain, logic hosted).  
+3. **Self-Service Portal (SSP)** — the real product surface for SMBs (see §5.2).  
+4. **Orchestration, not reinvention** — Cal.com (booking MCP already exists), Stripe (payments). We do not replace them.  
+5. **Telemetry + optimization reports** — “how often AI hit you, what they asked, what your site failed to answer, what local demand looks like.”  
+6. **Partner / intermediate surfaces** — agencies, SEO shops, tourism boards, chambers.
+
+### 5.2 Self-Service Portal (SSP) — primary product experience
+
+From the founder narrative (canonical UX):
+
+1. Business lands on geolocal.io, describes business → **genre-aware onboarding** (auto repair vs surf shop vs charter).  
+2. **Initial site scrape** → reflect back what the system already “sees.”  
+3. **Live “phone + chat” simulation** — show ChatGPT-style flow: recommend the business, list services, book a time (e.g. Thursday 3pm transmission filter). Instant aha: *pre-qualified, scoped, calendar-aligned demand*.  
+4. Install **tiny pointer** on site.  
+5. **Preflight / test run** in-browser against *their* MCP — if AI can’t see services, they fix content or portal data before going live.  
+6. **Upstream guidance** (discovery: GBP, NAP, FAQ, plain-text city+service) + **downstream guidance** (MCP content quality).  
+7. Optional **partner referral** for website help.  
+8. **Weekly/monthly reports** — interaction counts, intents, local demand signals.
+
+**Critical clarification (founder):** We are **not** selling them a chatbot for their customers. The portal’s HTML chat is a **test harness** wired to their MCP so owners can see how *external* AIs will interpret them.
+
+### 5.3 Genre-specific primitives
+
+Generic `get_business_info` is MVP scaffolding only. Differentiator is **genre systems**:
+
+- Auto: makes/models, specialties (Korean transmissions), emergency flags  
+- Beauty: services, duration, stylist notes  
+- Home services: service radius, emergency, estimate vs fixed price  
+- Tourism activities: seasonality, capacity, weather sensitivity  
+- Intermediate (tourism board): member directory + category routing
+
+Roadmap must version **genre packs**, not only generic tools.
+
+### 5.4 What we explicitly do NOT build (V1–V2)
+
+- Consumer destination app  
+- Competing with Calendly/Cal.com/Square booking engines  
+- Competing with Stripe/Square payments  
+- Replacing Google Business Profile  
+- Building a Yelp review network  
+- Boiling the ocean on travel (AirBNB/Kayak space) — stay complementary
+
+---
+
+## 6. Positioning
+
+### 6.1 Category
+
+**AI-readiness infrastructure / agentic local commerce infrastructure**
+
+Analogies that work in sales:
+
+| Analogy | Why it lands |
+|---------|----------------|
+| **Shopify Storefront MCP for local services** | Best single analogy; founders and partners get it immediately |
+| **Stripe for AI discovery/booking presence** | Infrastructure, not UI |
+| **Twilio for local service tools** | Agents call APIs; business doesn’t rebuild stack |
+
+### 6.2 Messaging (use this)
+
+- **Owner:** “Make every AI assistant understand and book your business.”  
+- **Tourism board:** “Turn lodging tax dollars into AI-discoverable local experiences your members own.”  
+- **Agency/partner:** “Add AI-readiness as a productized line — not another chatbot retainer.”  
+- **Investor:** “Neutral MCP layer for the 98%+ of local services AI currently skips, sold via self-serve + municipal distribution.”
+
+### 6.3 Anti-messaging (do not say)
+
+- “AI chatbot for your website”  
+- “Replace Google / Yelp”  
+- “SEO tool with AI features” (we can *include* discovery guidance, but category is infrastructure)
+
+---
+
+## 7. Go-to-market (refined)
+
+### 7.1 Primary V1 wedge: tourism / destination intermediaries
+
+**Why (copilot + market logic):**
+
+- Hotel/lodging taxes create **mandated promotion budgets**.  
+- Boards often **must spend** and openly solicit ideas — confused buyers with money.  
+- One intermediate win → **dozens/hundreds of endpoints** (members).  
+- Travel has high AI referral growth narrative (industry reporting of strong AI travel referral growth 2024–2025).  
+- Physical brochure racks prove the *intent* for local discovery; geolocal is the AI-native equivalent.
+
+**Motion:**
+
+1. Pick **1–2 destination markets** (illustrative: coastal FL tourism board style markets already used in founder narrative).  
+2. Sell **board-level pilot**: destination MCP + member onboarding SSP + simple ROI dashboard (member AI hits, category demand).  
+3. Fund via tourism marketing budgets / innovation RFPs / “AI visitor experience” framing.  
+4. Members get free/discounted endpoint for pilot period → convert to paid.
+
+### 7.2 Parallel motion: self-service SMB endpoint
+
+- Verticals for first genre packs: **auto repair, beauty, home services** (home services align with high-intent AI queries; tourism activities if destination wedge).  
+- Price band (direction from copilot pricing discussion, infrastructure framing):  
+  - **Starter ~$49/mo**  
+  - **Core ~$129/mo (anchor)**  
+  - **Pro ~$249/mo**  
+- Position above DIY “chatbot” tools; below enterprise multi-location AI visibility platforms (SOCi class).
+
+### 7.3 Secondary motion: partners (agencies, SEO, web shops)
+
+- Not the *product* owner narrative; they are **distribution** and **implementation**.  
+- Partner marketplace when SSP shows “your website needs work — self-serve or partner?”  
+- Commission / wholesale pricing TBD after first 50–100 endpoints.
+
+### 7.4 Explicit non-wedge
+
+- Do not lead with “compete in national multi-location brand AI visibility” (SOCi’s world).  
+- Do not lead with pure e-commerce (Shopify already owns that MCP).  
+- Do not lead with pure B2B / non-local.
+
+---
+
+## 8. Competitive strategy
+
+### 8.1 Who fills the vacuum?
+
+| Actor | Likely path | Our response |
+|-------|-------------|--------------|
+| **Yelp** | Become default AI source of truth via MCP + licensing | Offer **business-owned** data + booking path Yelp doesn’t control; complementary where agents need owner truth |
+| **Google** | Maps + Merchant + UCP | Stay complementary; deep GBP guidance; don’t fight Maps for UI |
+| **Shopify** | Expand agentic commerce | Partner pattern, don’t compete; copy their **well-known endpoint on merchant domain** UX |
+| **Cal.com / Square / Vagaro** | Booking MCPs | Integrate; own discovery + genre + intermediate + graph |
+| **AI visibility SaaS (SOCi, BrightLocal AI tools)** | Measure/optimize multi-location | Different buyer (enterprise brands); we own long-tail endpoints + tourism |
+| **Local agencies** | Manual AEO/GEO retainers | Make them partners; productize what they can’t scale |
+
+### 8.2 Moats (in order of build)
+
+1. **Endpoint density** in target geos/genres (critical mass for Model 3–4)  
+2. **Genre primitive quality** AI engines prefer (helpfulness repeatability)  
+3. **Telemetry flywheel** (demand signals → better recommendations → more attach)  
+4. **Intermediate contracts** (tourism/chambers lock distribution)  
+5. **Trust/quality enforcement** (decommission bad actors; freshness)
+
+Freshness moat vs Foursquare: demand is continuous booking/intent from the service economy, not social check-ins.
+
+---
+
+## 9. Pricing principles (SMB self-serve)
+
+| Tier | Monthly | Includes (V1 intent) |
+|------|---------|----------------------|
+| Starter | ~$49 | MCP endpoint, SSP test page, monthly report, basic diagnostics |
+| Core | ~$129 | Full MCP, live SSP preflight, weekly reports, local demand insights, partner access |
+| Pro | ~$249 | Multi-site, advanced telemetry, competitor signals, priority support |
+
+**Add-ons later:** partner install, advanced analytics, multi-location.  
+**Tourism board:** separate enterprise/pilot pricing (per-destination + per-member), not SMB list price.
+
+**Do not** price as chatbot usage; price as **infrastructure + BI**.
+
+---
+
+## 10. Success metrics (what GM tracks)
+
+### 10.1 90 days
+
+| Metric | Target (planning bar) |
+|--------|------------------------|
+| Public multi-tenant MCP (HTTP) live | Yes |
+| SSP MVP (onboard + scrape reflection + preflight sim + pointer install) | Yes |
+| Live endpoints | 100 |
+| Tourism board / destination pilots | 1 signed LOI or pilot |
+| Genre packs | 1–2 (e.g. auto + tourism activity) |
+| Booking path | Cal.com link or Cal.com MCP orchestration (not rebuild) |
+| Paying | First $ ARR even if small — proves willingness |
+
+### 10.2 180 days
+
+| Metric | Target |
+|--------|--------|
+| Endpoints | 1,000 |
+| Genres | 3 |
+| Related-businesses handshake in production | Yes |
+| Intermediate dashboards | 3+ boards/chambers active |
+| Agent telemetry usable for reports | Yes |
+
+### 10.3 360 days
+
+| Metric | Target |
+|--------|--------|
+| Endpoints | 10,000 |
+| Major CoC / tourism partnerships | 3+ |
+| Early DaaS / anonymized demand product | In market or beta |
+| Evidence AI engines prefer geolocal paths in pilot geos | Case studies |
+
+---
+
+## 11. Product / technical priorities (strategy-level only)
+
+Aligned to copilot “user stories before architecture”:
+
+**P0 — Prove Model 1**
+
+1. HTTP MCP transport (agents are remote; stdio is not the product)  
+2. Multi-tenant routing by business slug/domain  
+3. SSP: signup → scrape → simulation → pointer → preflight  
+4. Manifest / well-known generator  
+5. Seed genres + real pilot businesses  
+6. Cal.com path (redirect first; API/MCP next)
+
+**P1 — Prove Model 2**
+
+7. Intermediate (tourism) MCP + member aggregation  
+8. Board dashboard: members, hits, categories  
+9. Pilot paperwork + data agreements
+
+**P2 — Feed Models 3–4**
+
+10. Telemetry pipeline + owner reports  
+11. Related businesses tool + quality rules  
+12. Genre pack expansion  
+13. Partner marketplace
+
+**Explicit defer:** white-label everything, OAuth complexity beyond need, public developer platform, full DaaS, WordPress plugins (until attach rate demands them).
+
+---
+
+## 12. Risks (steelman)
+
+| Risk | Severity | Mitigation |
+|------|----------|------------|
+| Yelp/Google become “good enough” AI local sources | High | Own business truth + booking + genre depth; intermediates; owner control narrative |
+| SMBs don’t care until bookings prove ROI | High | SSP aha demo; tourism funded pilot; reports show demand; partner installs |
+| MCP / discovery standards shift | Medium | Follow Shopify/Yelp well-known patterns; multi-path discovery |
+| Market stats overstated in pitch | Medium | Cite SOCi/BrightLocal carefully; separate “multi-location study” vs “all SMBs” |
+| Scope creep into chatbot / marketplace | High | North Star discipline; weekly “are we infrastructure?” review |
+| Technical overbuild before SSP | High | No new tools without SSP path to value |
+| Tourism procurement slow | Medium | Parallel self-serve SMB + short pilot contracts |
+
+---
+
+## 13. Document governance (end the fracture)
+
+| Document | Role |
+|----------|------|
+| `docs/geolocal-copilot-conversation.md` | Foundational founder narrative (historical truth) |
+| **`docs/strategy/CANONICAL_STRATEGY.md` (this file)** | **Operating strategy — team rallies here** |
+| `NORTH_STAR.md` | Short public/internal compass — must match this file |
+| `docs/gtm/*` | Execution detail — must not invent a different company |
+| `docs/investors/*` | External packaging of *this* strategy |
+| `docs/engineering/*` | Build plan for P0–P2 above |
+| `code/` | Implementation of P0 |
+
+**Rule:** If a doc conflicts with this file, this file wins until the GM revises it.
+
+---
+
+## 14. Near-term decisions for the leadership team
+
+1. **Adopt this document** as the single operating strategy.  
+2. **Rewrite NORTH_STAR.md** to include four models + tourism wedge + SSP (not partner-dashboard-first only).  
+3. **Retire or mark “stale”** any GTM doc that makes partner dashboard the only product or collapses four models to one.  
+4. **Lock V1 markets:** choose destination pilot shortlist + 1 SMB vertical for genre pack.  
+5. **Lock pricing experiments:** Core $129 anchor A/B later; don’t block launch on pricing perfection.  
+6. **Engineering charter:** 30-day sprint = HTTP MCP + SSP spine + 10 real pilot endpoints (not more roadmap prose).  
+7. **Legal:** tourism pilot data agreement + business ToS + MCP telemetry privacy (lightweight now, not full DaaS).
+
+---
+
+## 15. GM summary
+
+**The copilot version is the right company.** The repo version got flattened into “partner MCP for SMBs with Cal.com” and lost the multi-model ambition, the SSP product, the tourism wedge, and the service-graph endgame.
+
+Market evidence in 2026 **supports** the urgency:
+
+- Consumers are using AI for local discovery at scale.  
+- AI recommendation is extremely selective.  
+- Shopify proved the pattern for *commerce* MCP on merchant domains.  
+- Yelp/Google/Cal.com prove MCP is real — and each protects *their* layer.  
+- Nobody is Shopify for local services’ AI presence.
+
+**Our job as a funded planning team:** freeze this strategy, realign all docs, and execute the SSP + tourism pilot path until the vacuum starts filling under *our* brand — not Yelp’s.
+
+---
+
+## Research sources (selected)
+
+- BrightLocal LCRS AI trust / local consumer research (2026) — consumer AI usage for local recommendations  
+- SOCi 2026 Local Visibility Index / PR coverage — 1.2% ChatGPT recommendation rate, selectivity vs Google 3-pack  
+- Shopify Storefront MCP docs / industry writeups — merchant-domain MCP, agentic commerce  
+- Yelp Fusion AI MCP (github.com/Yelp/yelp-mcp) — official local data MCP  
+- Google Cloud / Merchant API MCP materials — Maps grounding, Merchant MCP  
+- Cal.com MCP docs — booking lifecycle tools for agents  
+- X discourse (2026) — agentic commerce, AI local discovery, SMB invisibility narrative  
+
+*Stats should be re-verified at each external pitch; do not invent precision beyond sources.*