Skip to content

HANDOFF.md

Latest commit

 

History

History
125 lines (101 loc) · 7.78 KB

HANDOFF.md

File metadata and controls

125 lines (101 loc) · 7.78 KB

Handoff — Growth + Conversion Session (2026-06-14)

Last updated: 2026-06-15 · Author: Claude Code (Opus 4.8) · Branch: main

Purpose: let the next session pick up seamlessly. Read this + docs/human-todo.md

  • docs/dev-plan-2026-05-31.md (the validated V2 strategy) + STRUCTURE.md first.

2026-06-15 update (remote session)

Confirmed with the owner before coding. Shipped the last two code-doable WTP probes; everything else is now genuinely validation + distribution + time.

  • Looked at real funnel data (first-party DB on the Droplet): pipeline is healthy (pageviews ~800–1000/day, track() events land), but the preorder funnel has 0 events / 0 reservations since the 06-14 launch — /preorder/ just gets no traffic (it's buried behind low-traffic pages). GSC at T+1 day cannot yet show the 474 word pages (indexing takes days–weeks; report lags 2–3 days), so no GSC read yet.
  • word-to-color WTP paywall — BUILT + ENABLED (owner: "建并直接开"). After 5 distinct word lookups the palette/exports gate behind a /pro/ CTA + email-unlock. SEO-safe (counts only new user-typed words; crawlers / shared ?q= links / the 474 static pages never hit it). Toggle: WORD_PAYWALL_ENABLED in word-color-generator-page.tsx. Events: word_paywall_hit / word_paywall_pro_click / word_paywall_email_unlock.
  • Pre-order card test — LIVE 2026-06-15. Store is JPY, so priced ¥4,999 founder / ¥9,999 regular (referenced the existing Pro scale). Owner created the LS one-time product; NEXT_PUBLIC_PREORDER_CHECKOUT_URL set in Vercel via vercel CLI → /preorder/ shows the card button (verified). Kill criterion: <10 real card pre-orders by ~2026-07-15 → stop building Pro (LS → Orders, live mode). Setup notes: docs/preorder-ls-setup-2026-06-15.md.
  • LS/Stripe cleanup: duplicate LS store "Color Archive" (empty) being deleted via hello@lemonsqueezy.com; the separate paused Stripe acct is unrelated to the live MoR flow.
  • Owner's immediate move (both have execution docs now, written 2026-06-15):
    1. Distribution — docs/distribution-plan-2026-06-15.md (14-day cadence, CTAs routed to the live paywall/preorder). This is what feeds the empty funnels.
    2. 10 user interviews — docs/user-interview-recruitment-2026-06-15.md (copy + channels + 7-day outreach). Set up a booking link, start with the 5 registered users. Then let GSC + the funnels mature ~2–3 weeks before the S2 exit-gate decision. Don't add features.

The 2026-06-14 notes below remain accurate as the underlying state.

TL;DR

A 7-deploy session pushed all the code-executable growth + conversion levers: exposure (SEO/GEO/internal links/474 word pages), Core Web Vitals, a backlink engine, conversion fixes, and a full willingness-to-pay (WTP) pre-order funnel. The bottleneck is now validation + distribution + time — not code. Per the original dual-reviewed audit, the next real signal comes from running the WTP test + user interviews + GSC maturing, not from more commits.

What shipped this session (7 deploys, newest first)

Commit Lever
5e5ac2c Auditor pre-order CTA on /palette-audit/ + /wcag-audit/ (attributed preorder_cta_click {from})
61c6454 WTP pre-order landing /preorder/ (Accessibility Auditor, $49 founder)
174d155 Conversion: removed fabricated testimonial → honest trust row; ¥→JP¥; email capture on /word-to-color/
436eb32 Backlink engine: static HTML color badge on color pages + embed landing; fixed broken ${SITE_URL} attribution link; /embed/embed-code/ into sitemap+footer
d3e483d Core Web Vitals: Sentry Replay off (~50KB), gtag lazyOnload, stale PH banner off
ee32bb6 Exposure 2: 474 /word-to-color/[word]/ pages + index hub, story/use-case OG, GEO llms.txt
19c8063 Exposure 1: color→guide internal links, dedup 21 guide slugs, guide FAQ+CTR titles, IndexNow, hero COTD fix

All deploys: typecheck + build green before push; one commit per batch; session lock released after each.

Current state / key facts

  • Traffic (baseline 2026-05-31): ~110–240 PV/day. #1 entry /word-to-color/; top guides = color-psychology. ChatGPT is the #2 external source (AI engines ≈ Google organic). GSC: ~26.4K impressions stuck at avg pos ~13 (page 2), 0.8% CTR.
  • Revenue: ~0. Nobody has ever paid for Pro. Core diagnosis = PMF/traffic mismatch (casual info-seekers vs a ¥499/mo designer subscription).
  • New surfaces: 474 word pages, color→guide links, guide FAQs, dynamic OG everywhere, IndexNow, color-badge backlink engine, /preorder/ WTP funnel.

🔴 NEXT STEPS — prioritized (mostly NOT code)

1. Human-only, unblocks everything (this week):

  • Turn the WTP test into a real card test: create a $49 Lemon Squeezy "Accessibility Auditor — Pre-order" product → set NEXT_PUBLIC_PREORDER_CHECKOUT_URL in Vercel Production → redeploy. Until then /preorder/ only collects email reservations (weak signal). Kill criterion: <10 real card pre-orders in 30 days → designer-Pro theory not validated; stop building Pro.
  • Reconcile pricing: priceUsd in src/lib/checkout-config.ts doesn't match the JPY amounts (¥499≈$3.34 vs $6.99). Decide real prices + whether to bill USD at all. (Code change is trivial once decided.)

2. Code replacements impossible (this month):

  • 10 user interviews — script at docs/user-interview-script.md. The only source of "who/why/why-pay". This is the V2 plan's core and still not started.
  • Post the distribution drafts — docs/backlink-distribution-drafts-2026-06-14.md (Dev.to / Show HN / Reddit / directories). Real new traffic + backlinks.

3. Measure (2–3 weeks), then decide at the V2 exit gate (dev-plan-2026-05-31.md §7):

  • GSC: are the 474 word pages indexed? did page-2→page-1 move? CTR on top guides?
  • Funnel (PostHog): preorder_cta_click {from} → preorder_view → preorder_checkout_clicked.

Code items available if the user wants them (lower priority / need a decision):

  • word-to-color paywall after N free uses (2nd WTP probe) — WILL cut traffic; user must opt in.
  • Draft the interview recruitment copy + channel list (pure output; offered, not yet done).
  • Edge exposure polish: more OG routes (notes/vs), more word seeds, palette embed badge on collections (builder buildPaletteBadgeHtml already exists).

Protocol reminders (from CLAUDE.md — follow exactly)

  • Before any work: git pull --rebase origin main; read .claude/session-lock.json; if locked by autopilot, wait. Acquire lock as "lockedBy":"remote" via Bash echo.
  • One commit + one push per batch (every push = a Vercel build of 4,400+ pages). vercel.json ignoreCommand skips docs/metadata-only pushes.
  • Validate with npm run typecheck + npm run build before pushing (no test suite).
  • Release the lock (null values) and include it in the single commit.
  • Language: Chinese or English only. Minimize manual steps for the user.
  • Update STRUCTURE.md when pages/components/data/routes change.

Key files

  • WTP: app/preorder/page.tsx, src/components/preorder-page.tsx, src/components/auditor-preorder-cta.tsx, preorderConfig in src/lib/checkout-config.ts
  • Word pages: app/word-to-color/[word]/, src/lib/word-to-color-seeds.ts
  • Internal links: src/lib/color-guide-links.ts; guide SEO: src/lib/guide-seo.ts
  • Backlinks: src/components/embed-badge-button.tsx, app/embed/embed-code/
  • IndexNow: scripts/indexnow-ping.mjs (postbuild), key public/c0107a3b9f2d4e8a8b6c1d5e7f0a2b34.txt
  • Strategy: docs/dev-plan-2026-05-31.md, docs/baseline-metrics-2026-05-31.md, docs/human-todo.md, docs/user-interview-script.md