11. 24-Month Gated Roadmap
There's no single target. The roadmap climbs the MRR ladder in WHAT-TO-DO-AND-AVOID.md one rung at a time. Each phase ends at a checkpoint where you decide whether to climb, hold (run it profitably at that level), or change course.
Rule: no later phase starts until the current checkpoint is hit. Days 1–90 are in 08-90-day-plan.md.
Phase 1: Prove people pay twice (Months 0–3)
| Track | Work |
|---|---|
| Customers | WhatsApp concierge for 50 payers, then the thin web app for 150 |
| Product | Speaking placement → 2D map with 5 work scenes × 3 levels → "scripted spine, live edges" scene engine → feedback in Hindi or Bengali → weekly speaking score and before/after clip → share card. 18+, AI disclosure, audio deletion and one-tap cancellation from day one. |
| Team | A paid spoken-English trainer (a few hours a month) who owns the rubric and spot-checks feedback |
| Checkpoint | 150 payers (~$375 MRR), ≥50% month-2 renewal, median speaking score +10% in 30 days, voice and model cost <20% of revenue |
Phase 2: Make it a habit (Months 4–6)
| Track | Work |
|---|---|
| Product | Role tracks (BPO voice, hospitality, retail, office and IT, interview prep). NPC mood variations. Mock interview (Pro). Scene memory. 30-day re-placement. Annual plans and minute top-ups. Android app via user-choice billing, if day-90 renewal held. |
| Distribution | A daily short-video routine generated from the scene library, scene challenge links, referral credits, and long-tail pages ("BPO interview questions in Bengali") |
| Checkpoint | $1k MRR, ≥25% of new users from shares or referrals, payer churn ≤12%/mo |
Phase 3: Add institutes (Months 7–12)
| Track | Work |
|---|---|
| Product | Seat licences and a trainer dashboard (levels, minutes, scores per learner). Batch placement report. Certificate of level (not presented as official accreditation). |
| Distribution | Pilots with 10 institutes: skilling centres, hotel-management and retail-training institutes, college placement cells, BPO recruiters. Aim for 5 paying at ₹149–249 per seat a month. Learners get a personal free month after the batch ends, which turns B2B into B2C. |
| Optional | A second explanation language by demand (e.g. Odia, Marathi, Telugu), since the scenes are shared. Test paid acquisition only in channels with CAC payback under 3 months. |
| Checkpoint | $2.5k MRR, ≥5 institutes paying, payer churn ≤12%/mo |
Phase 4: Scale what works (Months 13–24)
| Track | Work |
|---|---|
| Product | Custom scenes for institute clients (priced separately). Email and chat-writing scenes. Self-hosted open speech models (AI4Bharat, MIT-licensed) to cut voice cost, if volume justifies it. |
| Experiments (only by demand) | Bengali-medium learners in Bangladesh (local pricing). A "Learn Bengali for Kolkata" or "Learn Hindi" track for movers, only if inbound requests are frequent. |
| Distribution | Double down on the single channel that produced the most payers (content or institutes) and cut the rest |
| Team | A part-time institute-sales and support person before any engineering hire |
| Checkpoints | Month 18: $5k MRR. Month 24: $8k MRR. $10k–20k is the stretch outcome, and usually needs profitable paid acquisition or a large institute network. |
What each level takes
Assumptions:
- blended ARPU about $2.50 a month per payer (₹199/mo and ₹1,199/yr mix plus some ₹399 Pro, at ₹88 = $1);
- payer churn about 12% a month;
- 2.5% of activated users paying.
The 2.5% figure is close to RevenueCat's Education download-to-trial rate (6.5%) multiplied by trial-to-paid (about 30–38%).
| MRR | Payers needed | New payers a month (to replace churn at steady state) | Activated new users a month | Or: with institute seats | What usually gets you there |
|---|---|---|---|---|---|
| $1k | ~400 | ~48 | ~1,900 | — | One working organic channel plus the concierge base |
| $3k | ~1,200 | ~145 | ~5,800 | ~800 individual + 4 institutes × 100 seats | Organic content, a working share loop and the first institutes |
| $5k | ~2,000 | ~240 | ~9,600 | ~1,200 individual + 8 institutes × 100 seats | Content, share loop and a steady institute pipeline |
| $10k | ~4,000 | ~480 | ~19,000 | ~2,200 individual + 18 institutes × 100 seats | All of the above plus one paid channel that pays back |
| $20k | ~8,000 | ~960 | ~38,000 | ~4,000 individual + 40 institutes × 100 seats | Profitable paid acquisition, or a large institute network |
| $50k | ~20,000 | ~2,400 | ~96,000 | — | A funded growth team. Not reachable organically at these price points. |
Three levers matter more than any feature:
- Churn. Cutting payer churn from 12% to 8% a month cuts the new payers needed by a third. The speaking score and re-placement are churn features.
- Institute seats. A 100-seat institute at ₹199 is worth ~80 individual payers, with no ad spend and lower churn during the batch.
- Voice cost per minute. At ₹0.5–1 a minute, gross margin stays above 70%. At premium real-time prices, every active learner loses money.
Budget guardrails
- Voice and model costs below 20% of revenue (target 15% by month 12). Cache NPC audio and scene scripts, cap live minutes, and route to the cheapest stack that meets the accuracy bar.
- No paid acquisition before month 6. After that, only channels with CAC payback under 3 months.
- Gross margin after payment fees, store fees, and voice and model costs ≥70%.