Skip to content

What's Missing — Honest Inventory ​

Historical inventory: 2026-05-30. See the 2026-10-26 production runbook for current engineering gates and remaining external acceptance requirements.

Grouped by when it'll bite us, not by category. Each line tells you who owns the fix and a rough effort estimate so you can prioritise. Things marked "You" are decisions or external accounts only you can handle; "Me" is engineering work; "Lawyer" is what it says.

The single biggest gap is not on this list. It's the missing merchant. Until a real Dhaka shopkeeper has used Jobab for a week, every item below is prioritised by guess.


🔴 Blocks the pilot starting ​

GapOwnerEffort
Meta Business Verification submittedYou2–6 week clock — start tonight
Public HTTPS URL for the webhook (ngrok in dev, deploy in prod)You5 min for ngrok
Privacy policy hosted at a real URL (markdown already drafted at docs/legal/privacy-policy.md)You + hosting1 hour
First real merchant scheduledYou1 week of outreach

Three of four are non-engineering. The engineering side is done for pilot-start purposes.


🟡 Will surface in pilot weeks 1–4 ​

These get real-merchant feedback only — engineer them only if the pilot reveals they matter. Don't pre-build.

GapWhy it'll come upEffort
Forgot-password flowSomeone forgets within week 1Me · 2 hours
Email verification on sign-upCurrently any "@" passesMe · 2 hours
Email service (Resend / Postmark) for invites + password resetsTokens-via-WhatsApp works for v0Me · 3 hours
Conversation assignee picker in the inbox UIBackend supports it; no UI dropdownMe · 1 hour
Customer-facing AI disclosure toggleFirst time a customer realises they're talking to AIMe · 1 hour
Bulk actions (multi-select, bulk-takeover, mark-all-read)When a merchant has 50 DMs queuedMe · half day
Full-text search across message contentCurrent search is name-onlyMe · 3 hours
Daily WhatsApp summary ("Jobab handled 47 today, ৳12k revenue")Without it merchants underestimate value deliveredMe · half day
AI-off scheduling ("pause AI 11pm–7am")Some merchants will want strict hoursMe · half day
Catalog editing in UI (single-product add/edit, stock adjust)Currently CSV-only — frustrating for one-off changesMe · 1 day
Order line-item editing post-creation"I forgot to add a dupatta"Me · half day
Refund / partial refund flowFirst real complaint comes, no UI to resolve cleanlyMe · 1 day
Date-range filtering in analytics + orders"Show me last week" — common askMe · 3 hours
CSV export of orders / messagesFor their accountantMe · 2 hours

🟠 Production launch blockers (before public sign-up) ​

Eventual must-haves. Pilot works without them.

GapOwnerEffort
Meta App Review approvedYouWeeks of process
Real bKash merchant accountYouBusiness onboarding
Stripe (or equivalent) for our billingYou + Me1–2 days
Pricing tiers + paywall + usage capsProduct + Me3–5 days
Trial mode with auto-expiryMe1 day
Per-org LLM cost caps (so a runaway customer can't drain margin)Mehalf day
MFA / SSO (replace dev auth with Clerk / Supabase)Me + You (account)1 day
Real Sentry + Langfuse keys in productionYou (signup)10 min once accounts exist
Terms of service + DPALawyer1–2 weeks
Cookie consent banner (if any EU traffic)Me2 hours
Data Processing Agreement template for design partnersLawyer1 week
Production hosting (Railway / Render / Vercel / Fly)You1 day
Database backup + point-in-time recoveryYou (hosting choice)depends

🔵 Phase 2 / scale items (don't build yet) ​

GapWhy deferred
Real Jina API key + visual ANN activeImage match works as describe-then-search; upgrade when volume justifies cost
Multi-page UI (org with multiple FB Pages)Schema supports it; no merchant needs it in pilot
Mobile image-attachment composerRead-only mobile is enough for pilot
Mobile push delivery tested on a real deviceCode is wired; needs Expo / EAS run
Mobile app icons + splash + EAS build setupPre-launch polish
Voice messagesOut of MVP scope
Reports / exports + per-product analytics"What's selling?" — useful, not load-bearing
Customer / CRM view ("all of Tahmina's history")Power feature; not pilot-critical
Tags / labels / notes on conversationsPower feature
Webhook-subscription management UI for multi-Page orgsPhase 2 §11 follow-on
Comments-page rule preview / sandboxCurrently merchant must PATCH and watch

🎨 UI polish still on the to-do list ​

GapEffort
Lucide icons everywhere (replace hand-drawn SVGs)1 hour — massive cohesion win
Type + spacing scale locked in tailwind.config.tshalf day
Per-screen empty-state illustrations (no longer one shared recipe)1 day with a designer
Port the inbox hero treatment to every page via AppShell2 hours
Compact list view for /orders when ≥20 ordershalf day
Storybook so a designer can iterate without an engineer1 day
Favicon, PWA manifest, OpenGraph card, login background pattern2 hours + designer for assets
404 illustration instead of text-only30 min + asset

🧪 Engineering rigor we haven't shipped ​

GapNote
Web unit tests0 today. Vitest + Testing Library — 1–2 days for inbox + auth coverage
Mobile tests0. Same Vitest setup
E2E tests (Playwright)1 day for the critical paths (sign-up → first AI reply, order → mark paid)
Integration tests against a real Postgres in CISchema is in CI already; just write the supertest cases
Performance budgetsLighthouse not run; bundle size not measured
Index tuningSensible indexes exist; no production-query analysis yet
Accessibility audit (axe + manual screen-reader pass)Not done
Backend deploy step in CIWorkflow runs lint / typecheck / test; no deploy stage
Real production Dockerfile runMulti-stage Dockerfiles exist; never run in prod context
Audit-event retention policyCurrently keeps forever
Encryption-key rotation runbookConceptual; no procedure documented

💰 Things only you can decide ​

These aren't engineering gaps — they're product/business decisions outstanding.

  1. Pricing model. 1% GMV? Flat $25/mo? Per-conversation? Decide before pilot week 3.
  2. Free trial length. 14 days? 30? Forever-free with a usage cap?
  3. Company legal entity. Where, what structure, who owns what.
  4. Domain name. jobab.com? Something else?
  5. Brand mark. Jamdani motif is good; want a real logo lock-up from a designer?
  6. Founding team structure. Solo? co-founders?
  7. Investor strategy. Bootstrap? Pre-seed? Grants from BD ecosystem (LightCastle, Startup Bangladesh)?
  8. Geographic scope. Dhaka first, then where?
  9. Vertical scope. Clothing only (current bias)? Food? Cosmetics? Phones?
  10. AI disclosure policy. Default on/off + the exact phrasing.
  11. Privacy policy + DPA reviewed by a lawyer in your jurisdiction.
  12. Meta Business Verification doc package assembled (we have the kit; need the legal-entity papers).

📊 Where we are vs. where the next bottleneck is ​

StageStatus
Engineering — Phase 1✅ done
Engineering — Phase 2✅ done
Engineering — UX polish from the review✅ shipped
Pilot prep — docs + eval kit✅ done
Pilot prep — Meta App Review materials✅ docs done · ❌ not submitted
First real merchant signed up❌ zero
First real customer DM served❌ zero (only fake-webhook traffic)
Bangla quality at production grade❌ unknown without eval data
Pricing decided❌
Billing infrastructure❌
Public launch🚧 6–12 weeks away minimum

My recommendation (PM hat) ​

Stop adding features. The pile of code is now larger than the pile of evidence that anyone wants it. The cheapest, fastest, most informative thing this week:

  1. Today — submit Meta Business Verification.
  2. Today — ngrok http 3000, paste the URL into the Meta App webhook + data-deletion config.
  3. Today — cold-DM 8 candidate merchants from docs/ship/pilot/pilot-outreach.md.
  4. This week — do 5 real merchant interviews. Don't pitch — listen (docs/ship/pilot/interview-script.md).
  5. Next week — install Jobab on one real merchant's Page. Watch it for 4 weeks.

Everything in the lists above gets re-prioritised by what merchants actually surface during those 4 weeks. Anything we'd build today without that signal is guessing — well-engineered guessing, but still guessing.


How to use this file ​

  • Re-read at the start of every planning session.
  • Move items between sections as priorities shift (pilot data should cause big movements).
  • Anything that gets done → strike it through, leave the line, add a date. That history becomes the project diary.
  • Add new lines as you discover them. The list is a living thing.

MIT licensed