# How this makes money (and why "charge 3¢ to add a cookbook" doesn't work)

**Plain-English money plan — 2026-06-09.** Companion to PLATFORM.md.

## The one thing that breaks naive micro-charges: the 30¢ floor

Every card payment costs **~2.9% + $0.30**, and the **30¢ is fixed per transaction.** That single fact decides the whole model:

| User loads… | You keep | Lost to fees |
|---|---|---|
| $1 | ~$0.67 | **33%** |
| $5 | ~$4.56 | 9% |
| $10 | ~$9.41 | 6% |
| a direct 3¢ charge | **−27¢** | impossible (fee > charge) |

So you can **never** charge per-action directly — you'd lose money on every click. Anything usage-based **must** be a prepaid balance, and you should nudge **$5+ top-ups** so the 30¢ stops hurting. But there's a bigger problem than fees…

## The real trap: charging for the thing you want to spread

Our whole edge is **virality through real networks** ("everyone's network's network"). If someone must enter a credit card to *add a cookbook*, they won't — and the network effect dies at step one. **Never paywall the action you want people to do freely.** Adding cookbooks costs you ~1¢ each; that's not a cost, it's the cheapest growth you'll ever buy.

## The model: free to add, money on the cart + premium

```
   ┌─────────────────────── FREE (the growth engine) ───────────────────────┐
   │  browse · cook-from-what-you-have · add cookbooks (paste/photo/YouTube) │
   │  your cost: ~1¢ per recipe parsed  ·  funded by ↓                       │
   └────────────────────────────────────────────────────────────────────────┘
                 │                                            │
        ┌────────▼─────────┐                        ┌─────────▼──────────┐
        │  🛒 THE CART      │  ← main revenue        │  ⭐ PREMIUM (later) │
        │  Instacart pays   │                        │  $3–5/mo: unlimited │
        │  YOU a commission │                        │  scans, household,  │
        │  on the order     │                        │  meal plan, no ads  │
        └───────────────────┘                        └────────────────────┘
```

1. **Adding cookbooks = free.** It's pennies and it's the fuel. 1,000 people × 10 recipes ≈ **$100 of AI** — buys your whole early network.
2. **Make money on the cart, not the recipes.** When someone sends their missing-items list to Instacart, **Instacart pays *you*** — published rates: **up to $10 per *new-customer* order (CPA)**, **up to 15%** (influencer tier); Developer-Platform partner rates are negotiated. The grocery commission literally funds the free AI parsing. (Amazon's grocery affiliate is only **1%** — Instacart is the real earner.)
3. **Premium later** ($3–5/mo, benchmarked vs Mealime/Samsung Food): unlimited photo scans, household/shared cookbooks, meal planning. Charged at a size where the 30¢ fee is negligible. This is supplement, not the engine.
4. **Prepaid credits only if needed** — e.g. a power user importing hundreds of recipes via AI. Sell "credits" (a product, not stored cash — simpler legally), $5 minimum.

## Worked example (rough, ~1,000 active users/month)

**Costs**
| | /mo |
|---|---|
| AI parsing (~10 adds/user × ~1¢) | ~$100 |
| Hosting (Cloudflare) | ~$5 |
| **Total** | **~$105** |

**Revenue (illustrative)** — say 5% of users place a *new-customer* Instacart order:
- 50 orders × ~$10 CPA = **~$500** → comfortably covers the ~$105 cost, with margin.
- (Repeat-order commissions are thinner; the new-customer CPA is the lucrative bucket. Real rates come only after you apply via Instacart's Impact program.)

Even at much lower conversion, the **cost side is tiny** — the risk isn't going broke on AI, it's not getting enough users/orders. Which is exactly why **free adding (for virality) beats charging to add.**

## Cost levers (keep the AI bill near zero)
- **Cache** parses — never parse the same cookbook/photo twice (Cloudflare KV).
- **Batch API** for non-urgent parses = **−50%**.
- **Cheapest model that's good enough** (Haiku / a free-tier model / your own 3090 for personal use).
- The free **blank-line paste** + **TheMealDB** + **URL import** add content at **$0** (no AI at all).

## Watch-outs
- **Stored value / wallets** can trigger money-transmitter rules + refund duties at scale. Frame paid balance as "credits" (a product), keep clear ToS, don't hold real cash floats. Trivial at small scale; plan for it if it grows.
- **Don't depend on one rail.** Instacart commissions are negotiated/revocable (platforms close APIs — Quizlet, Netflix). Keep the cart layer swappable (Walmart/Kroger/Amazon/plain copy) so commissions are upside, not a single point of failure.
- **Affiliate income only compounds at scale** — early on it won't pay the bills. Keep costs near zero (you can) and treat revenue as a later-stage problem.

## TL;DR
Give away the **adding** (pennies, and it's your growth). Earn on the **cart** (Instacart pays you) and a small **premium** tier. Any usage charge must be **prepaid $5+** because of the 30¢-per-transaction floor. The business is viable *because the costs are tiny* — so optimize for adoption/virality first, monetization second.