# P13 — Pricing & Unit Economics

**Updated:** 30 August 2026 · Platform prices web-verified 30 Aug 2026 (sources: `FOUNDER_GUIDE.md` §8). Re-verify on vendor pages before quoting — third-party round-ups drift. All labour figures are estimates until pilot data replaces them (`PROGRESS.md`).

---

## 1. The tiers (portfolio-bound: $99–199/mo, setup included)

| | **Starter — $99/mo** | **Growth — $149/mo** | **Pro — $199/mo** |
|---|---|---|---|
| Included AI minutes/mo | **250** | **500** | **800** |
| Overage | $0.25/min | $0.25/min | $0.22/min |
| Channels | Phone OR WhatsApp | Phone + WhatsApp | Phone + WhatsApp |
| Booking integration | 1 calendar | 1 calendar | Calendar + CRM webhook |
| Voice | Standard | Standard | Premium voice option |
| Reporting | Daily summary | Daily summary + weekly report | + monthly review call |
| Escalation to human | ✅ always (D8 — not a tier feature) | ✅ | ✅ |
| Setup | Included | Included | Included |
| Target buyer | Salon, small clinic | Restaurant, busy clinic | Real-estate, multi-staff clinic |

**No unlimited plan exists at any price** (`SPEC.md` D7). ~250 min ≈ 8–10 short calls/day; ~800 min ≈ 25–30. A client who consistently overruns is a *success story* — the upsell script is in `SALES_PLAYBOOK.md` §6.

Rules baked into every agreement: minute cap + overage stated on the invoice; month-to-month, cancel anytime (pilot honesty — retention must be earned, not contracted); price review allowed at 60 days if usage profile differs >2× from the quote.

## 2. Researched per-minute cost anatomy (the input data)

| Component | Budget stack | Premium stack | Source basis |
|---|---|---|---|
| Retell voice engine (STT+TTS+orchestration) | $0.070 | $0.070–0.080 | Retell pricing round-ups, Aug 2026 **[verify]** |
| LLM | $0.006 (4o-mini-class) | $0.03–0.16 (frontier-class) | Retell/Vapi component breakdowns |
| Telephony (US, via platform) | $0.015 | $0.015 | Retell docs ($0.015/min + $2/mo number) |
| **All-in $/min** | **≈ $0.091** | **≈ $0.12–0.26** | |
| Vapi alternative (BYO keys, budget) | ≈ $0.08–0.10 | ≈ $0.15–0.25 | Vapi $0.05 orchestration + components |
| Planning number used below | **$0.10/min** (deliberately above budget-stack estimate) | | |
| **MEASURED $/min (task `F-02`)** | **⛔ NOT YET MEASURED** | | Blocked on `F-01` → O2/O4. Instrument ready: `runbooks/R2-cost-measurement.md` + `runbooks/cost_calc.py` |

> **The $0.10 planning number is still an assumption, and every figure in §3 inherits that.** `PROGRESS.md` records the finding that third-party round-ups disagree across a **4× spread ($0.07–0.31/min)** — at the low end a Starter client costs $21.50/month in infrastructure; at the high end the same client costs $82.50 and is underwater before any labour. Replacing this row with a measured figure is task `F-02`; it takes 3 hours and ten phone calls, and it cannot be done from documentation. Until then, **do not quote a client a tier on the strength of the table below.**

Fixed per client: number $2–5/mo; WhatsApp Business Calling inbound = free (Meta) **[verify at BD/Gulf pilot]**; glue automation (n8n/Make) ≈ $0–20/mo across ALL clients.

## 3. Unit economics per tier (the honest math)

Planning number $0.10/min; assumes client uses 100% of included minutes (worst normal case):

| | Starter $99 | Growth $149 | Pro $199 (premium voice ≈ $0.16/min) |
|---|---|---|---|
| AI minutes cost | 250 × 0.10 = **$25.00** | 500 × 0.10 = **$50.00** | 800 × 0.16 = **$128.00** ⚠ |
| Pro on budget-stack voice instead | — | — | 800 × 0.10 = **$80.00** |
| Number + misc | $4 | $4 | $5 |
| **Infra COGS** | **$29** | **$54** | **$85–133** |
| **Infra % of fee** | **29%** ✅ | **36%** ✅ | **43–67%** ⚠/❌ |
| Gross margin before labour | $70 | $95 | $66–114 |
| Labour at target 2h/mo (valued $15/h) | $30 | $30 | $30 |
| **Net contribution/client/mo** | **≈ $40** | **≈ $65** | **≈ $36–84** |

**Verdict, said loudly:** $99–199/mo **is viable — but only capped**. Two specific consequences:

1. **Pro tier cannot include premium voice at full 800-minute usage** — premium voice at $0.16/min on 800 min breaches the 40% floor. Rule: Pro ships on the budget stack by default; premium voice is offered only where measured usage ≤ ~500 min/mo, or at +$29/mo add-on. This is a research finding, not a style choice.

> ### ⚠ Open finding, 30 Aug 2026 — the stated fix for Pro does not actually clear the floor
>
> Consequence 1 above prescribes "Pro ships on the budget stack by default". **That case is already in the table and it fails:** budget-stack Pro is $80 of minutes + $5 fixed = **$85 infra on a $199 fee = 42.7%**, against a 40% floor. The table's own verdict row says `43–67% ⚠/❌` — the ⚠ end *is* the budget stack. The remedy and the breach are the same number.
>
> At the $0.10 planning rate, the floor implies a Pro cap of **746 minutes**, not 800. (Verify: `python3 runbooks/cost_calc.py --rate 0.10`.)
>
> **How much this matters depends on a number we don't have.** The floor is checked monthly against **actual** usage, not the cap — a Pro client at 600 minutes sits at 33% and is fine. The defect is that the tier is *designed* so a fully-utilising Pro client is out of compliance on day one, which makes the floor unenforceable exactly where it was meant to bite. And if `F-02` measures $0.085/min, Pro at 800 lands at 36.7% and the whole question disappears.
>
> **Founder decision, after `F-02` — not before:** (a) cut the Pro cap to the floor-compliant figure · (b) price Pro at ~$215 · (c) state explicitly that the floor governs actual usage and that a capped-out Pro client triggers a re-tier conversation by design. Option (c) costs nothing and may well be what was always intended — but it needs writing down, because right now the document says one thing and its arithmetic says another.
2. An **uncapped** $99 client at restaurant-level volume (750+ min/mo) costs $75+ in infrastructure = margin ≤ 24% before any labour. This is why competitors bill per call (Smith.ai $1.60–1.90/call) or charge $399+ (Slang.ai). Caps are existential, not fine print.

### Margin floor rule (binding)

> **Infrastructure ≤ 40% of the client's monthly fee.** Checked monthly per client from the platform usage dashboard (`TESTING.md` §3). Two consecutive breach months → move the client up a tier or re-quote; never absorb silently.

### The labour kill line (verbatim from the portfolio)

> **Kill criteria: delivery time >10h/client/mo (service margin death) → drop or productize config into templates.**

At $99/mo, 10 h/mo = $9.90/h — below any alternative use of founder time. Hours are logged per client in `PROGRESS.md`; the template library (`SPEC.md` D4) is the standing mechanism that keeps steady-state care at the ≤ 2h/mo target.

## 4. Setup-fee decision

**Pilot phase (P1–P8): no separate setup fee** — the portfolio entry says "including setup" and pilot friction must be zero.
**Post-P8, new clients:** introduce a **$149 one-time setup fee** (waived on annual prepay). Rationale: templated setup will cost ~8h; $149 filters non-serious buyers without deterring real ones, and matches market practice (managed competitors charge $0–500 setup). Decision confirmed/rejected at the P8 gate with real setup-hour data.

## 5. Revenue sketch (estimates, not forecasts)

| Scenario (by ~P8, end Oct 2026) | MRR | 90-day cumulative revenue |
|---|---|---|
| Floor: 2 retained of 3 pilots (success bar) | 2×$99–149 ≈ **$200–300** | ≈ $500–900 |
| Base: 3 retained + 1 referral | ≈ **$400–550** | ≈ $900–1,400 |
| Stretch: + 2 US clients via LLC at $149–199 | ≈ **$700–950** | ≈ $1,300–2,200 |

Small absolute numbers — the pilot's value is (a) proving retention, (b) the template library, (c) a wedge into US SMB sales via the LLC where the same service sells at 1.5–2× BD willingness-to-pay with easier telephony. Scale math for later: 20 retained clients ≈ $2,500–3,500 MRR at ~35h/mo total care — only sane if templates hold the 2h line.

## 6. Comparison anchor (what the same thing costs elsewhere)

| Provider | Monthly price | Our position |
|---|---|---|
| Smith.ai (AI + human backup) | $95–800/mo + per-call | We undercut mid-tiers with more hands-on config |
| Slang.ai (restaurants) | $399–599/mo | We are 1/3 the price for the pilot verticals |
| Goodcall | from $79/mo | DIY-flavoured; we win on done-for-you + integrations + local language/channel (WhatsApp) |
| Rosie | from $49/mo | Same — price is not our battlefield; managed quality is |
| DIY on Vapi/Retell | ~$25–80/mo infra | The client would need to be their own prompt engineer — that gap IS our product |

Full landscape: `COMPETITOR_ANALYSIS.md`.