# Audit Gaps — Pending (from Yannis-persona critique)

**Status**: ~~empty~~ → ready for Phase B fill-in. Each item = 1-2 paragraphs in the relevant audit file, or a short new section.

## P0 — must fill before research.html ships

- [ ] **Business model math** — LTV, CAC, ARPU per market, churn curve. Novakid claims ~940K parents, 80M+ revenue. Work out rough unit economics. *Lives in*: new section in `platform-current-state.md` or standalone `business-model-math.md`.

- [ ] **Cohort / churn analysis** — When exactly do parents churn (month 1 / after free trial / post first slow-progress week)? This is THE churn signal the Value Hub targets. *Lives in*: `trustpilot-systematic.md` (already has directional Trustpilot data).

- [ ] **Renewal mechanics** — How auto-renew works, when dashboard surfaces it, how many days before charge. Direct retention lever. *Lives in*: `platform-current-state.md`.

- [ ] **Plateau / bad-week handling** — How current Novakid handles "child is struggling" or "child plateaued". Probably: it doesn't. But we must confirm, not guess. *Lives in*: `heuristic-evaluation.md` as a 7th lens or addendum.

- [ ] **Mobile vs desktop split** — Brief says desktop-only deliverable, but ~70%+ of parent opens are likely mobile. Acknowledge explicitly in audit + explain why desktop is still the right canvas for this exercise. *Lives in*: `heuristic-evaluation.md` preface.

## P1 — good-to-have

- [ ] **Teacher UX** — Teacher dashboard, Heart System visualized, what the teacher sees post-lesson. Partially covered in `trustpilot-systematic.md` (teacher perspective). Fill in: teacher side of the comment input flow. *Lives in*: new `teacher-ux-sidebar.md` or append to `heuristic-evaluation.md`.

- [ ] **AI narrative technical feasibility** — Can Novakid actually generate per-child weekly narratives? Cost, latency, batch vs real-time, data consent. If NO — our thesis weakens. *Lives in*: `platform-current-state.md` technical section.

- [ ] **Data infrastructure reality** — The claim "1,500 micro-skills tracked by AI" — is that verifiable? PR-speak or real? If we can't verify, we flag it. *Lives in*: `quote-verification.md` (extend to verify claims, not just quotes).

- [ ] **A/B test proxies** — What metric would prove the redesign worked? Weekly active dashboard open rate? Time-to-first-progress-share? Parent NPS at month 3? *Lives in*: `thesis-and-pov.md` or new `validation-plan.md` in `03-strategy/`.

- [ ] **Notification system audit** — Dashboard is the *pull* surface. What's the *push* system (emails, in-app, WhatsApp)? How do they couple? *Lives in*: `heuristic-evaluation.md`.

- [ ] **Payments/refund flow** — Biggest anxiety trigger (Trustpilot #1 complaint). Current refund path. *Lives in*: `heuristic-evaluation.md`.

- [ ] **Localization debt** — Turkish locale leakage into Speaking practice etc. QA signal + trust issue. *Lives in*: `heuristic-evaluation.md`.

---

**Principle**: each gap = max 1-2 paragraphs. No net-new deep research. The point is acknowledgment + directional evidence, not perfection.