Skip to content

ROADMAP_V3.md

Latest commit

 

History

History
84 lines (54 loc) · 5.93 KB

ROADMAP_V3.md

File metadata and controls

84 lines (54 loc) · 5.93 KB
Raw

🗺️ Classic Watch Pro — Product Roadmap (living document)

Reality-checked: every release below is tracked against working code, not doc claims. Latest shipped: v3.1.0 (trust & commerce + discovery + admin depth + E2E/telemetry). Companion docs: ../README.md, UPGRADE_SUMMARY.md, DEPLOYMENT.md.

✅ Release status

Release Theme Status
2.1 Baseline completion (tooling, UI layer, PWA, tests) ✅ Shipped (v2.1)
3.0-a Trust & commerce — front-end slice (reviews, promos, compare, notifications, refunds) ✅ Shipped (v3.1)
3.0-b Trust & commerce — backend slice (live payments, email/SMS, real backend) ⬜ Pending backend repo + credentials
3.1 Intelligence lite: recently-viewed, notifications, admin analytics depth ✅ Shipped (v3.1)
— Design refresh (ivory/navy palette, serif display face, editorial cards) ✅ Shipped (v3.1)
— QA & hardening: axe suite, Lighthouse CI budgets, Playwright E2E, Sentry ✅ Shipped (v3.1)
3.2 Semantic search & recommendations v2 (embeddings) ⬜ Depends on real data/backend
3.3 LLM concierge, review intelligence, admin/content copilots ⬜ Depends on 3.2 + RAG backend

🚢 v3.1.0 (September 2026) — what actually shipped

Trust & commerce (customer-facing)

  • Reviews & ratings: per-product list with rating-distribution summary, verified-purchase badges, helpful votes, and a write-review dialog. Mock DB gained a seeded reviews collection; product rating/reviewCount aggregates are recomputed from live rows on every write/delete, so the catalog never contradicts the review list.
  • Promo codes: WELCOME10 / LUXE200 / SUMMER20 seeded in the mock DB (percent + fixed, min-order, usage caps, expiry). Cart stores the applied promo; checkout shows the discount line, and order placement redeems it and records discount + promoCode on the order.
  • Notification center: header bell with unread badge + feed; order placement, status changes, refunds and cancellations push live notifications (toasts too).
  • Recently-viewed rail on Home (persisted, deduped, capped) and quick-view + compare drawer (up to 4 watches, spec-by-spec) on every product card.

Admin depth

  • One-click refunds for paid orders (immutable audit event + notification; ineligible orders are rejected).
  • Top customers by lifetime spend on the overview.
  • Review moderation queue (/admin/reviews) — search + verified badge + remove (recomputes aggregates).
  • Low-stock inventory panel (shipped earlier) kept; stock, orders and most-reviewed panels intact.

Design & performance/a11y (no regression)

  • Token-level refresh: warm ivory light background, midnight-navy primary, editorial Playfair Display headings (async, fixed metrics → CLS-safe swap). Dark mode palette untouched (still AA+).
  • 19-page axe regression suite (structural WCAG A/AA) incl. the new reviews page; Lighthouse CI budgets unchanged and green.

QA & release engineering

  • Playwright E2E job in CI: buy-flow (incl. promo apply), auth + admin guard, moderation smoke — against the production build.
  • Sentry wired but strictly lazy: zero bytes unless VITE_SENTRY_DSN is configured.
  • 152 unit/component tests (23 files); lint 0 / tsc 0; npm audit --omit=dev gate kept.

Demo-data model note: reviews and promoCodes are persisted in the schema-versioned envelope and hydrate deterministic seeds when a snapshot predates them — old localStorage/backups upgrade in place, never silently discard.

🔭 Remaining by theme

Release 3.0-b — Trust & commerce, backend half (needs the server repo + credentials)

  • Live payments (Stripe or regional rails) + webhook-driven paymentStatus; real email on status change; saved address book; guest checkout; authenticity/warranty/returns pages; gift cards.
  • Cross-device accounts once Firestore replaces the demo adapter.

Hardening track (continuous)

  • CI is green on every PR (3 jobs). Open items: preview deploys, Firestore security rules for the real backend, structured logging, checkout load test at 2× traffic.

Release 3.2 — Semantic search & recommendations v2

  • Embed product specifications/descriptions into a vector index; hybrid behavioral + embedding recommendations. Gates on real analytics events + backend.

Release 3.3 — LLM concierge & copilots

  • RAG concierge over the catalog/FAQ corpus (grounded answers + deep links, hallucination eval in CI), review pros/cons + sentiment summaries (structured review tags are captured today in Review), admin weekly summary, EN/BN content drafts. Depends on 3.2 + a backend RAG endpoint.

♟️ Sequencing principles

  1. Trust first, AI second — every AI phase compounds once orders/views/reviews are real; the data readiness prerequisite (clean taxonomy, event capture) is still the gating step.
  2. Demo parity — anything built client-side keeps working in zero-config demo mode; anything needing a backend is scoped separately and clearly flagged.
  3. Quality gates are non-negotiable — lint/tsc/tests/build/E2E/Lighthouse/audit all green before merge; a11y (axe) and Lighthouse budgets guard every PR.

📊 North-star metrics

Repeat-buyer + referred revenue share; conversion/AOV/cart-abandonment once live analytics exist; recommendation CTR ≥ 8% (3.2); concierge deflection ≥ 30% with 0 high-severity hallucination incidents (3.3).