Fig. A — full disclosure
Methods & assumptions
The landing page says the assumptions are on the table. This is the table. If you think a number below is wrong, email me — I'll either defend it with a reason or change it and note the change here.
A.1 — what one run does
The simulation loop
Each of the 2,000 runs simulates 36 months. Every run first draws its own version of your inputs from noisy distributions (below), because your guesses are guesses. Then, month by month: new customers arrive, a fraction churns, and — the part most models skip — customers you don't have the hours to support mostly leave. The result is that run's ARR at month 36. The histogram is all 2,000 results; the spaghetti chart shows 60 of the paths.
A.2 — distributions
How your inputs get noised
| Input | Distribution | Spread (σ) | Why |
|---|
| Monthly churn | lognormal around your guess, clamped 0.2%–60% | 0.45 | Founders' churn guesses are usually off by half; lognormal keeps it positive and right-skewed like real churn data. |
| Acquisition / month | lognormal around your guess | 0.50 | Acquisition is the least predictable input, so it gets the widest spread. |
| Hours / customer | lognormal around your guess | 0.35 | Support load varies less than growth, but heavy customers exist. |
| Price | fixed at your input | — | Price is the one thing you fully control, so it isn't noised. The lever table varies it for you. |
Lognormal draws use mean × exp(σZ − σ²/2) so the mean of the distribution equals your input rather than drifting above it.
A.3 — constraints
The rules that kill runs
- Support capacity. Only 60% of your weekly hours can go to supporting customers; the rest is assumed to go to building, selling, and admin. Capacity =
(weekly hours × 4.3 × 0.6) ÷ hours per customer.
- Overflow churn. When customers exceed capacity, 80% of the overflow leaves that month. Under-supported customers don't wait around.
- Selling needs hours too. Acquisition is scaled by
min(1, weekly hours ÷ 8) — below 8 hours a week, you win proportionally fewer customers.
- The cliff. From the month you set, your weekly hours drop to the after-cliff value and both rules above bite harder. This is the constraint that separates Breaks First from spreadsheet math.
A.4 — limitations
What the model does not know
- Whether anyone wants your product. The model takes your acquisition guess as given. If nobody wants the thing, no simulation saves you — do a one-hour competitor check first.
- Price elasticity. The "Price ×1.5" lever holds acquisition and churn constant. In reality raising price usually costs you some buyers — the lever shows the mechanical upside, and it's on you to judge how much of it survives contact with customers.
- Competition, platform risk, black swans. Not modelled. A competitor launching free, an API you depend on changing terms — none of that is in here.
- Compounding advantages. Word of mouth, SEO, and content that snowballs are also not modelled, so long-run good cases may be understated for products that earn distribution.
Breaks First is a stress-test, not a forecast. It's for killing fragile ideas cheaply, not for predicting the future of good ones.
A.5 — the boring but binding part
Refunds & licence
- Refunds: full refund within 7 days of purchase, no questions asked, no exit survey. Email the address below with your order number.
- Founder licence: one payment, one licence key, valid for this product and every future feature added to it, forever. Personal use on any of your own devices.
- Payments & keys: handled by Lemon Squeezy (the merchant of record). Your key arrives in their receipt email; paste it into the unlock box on the instrument.
- Lost key: email me, I'll resend it.
- Data: your simulation inputs never leave your browser. There is no backend, no analytics on your numbers, and no account. The only network call the instrument makes is licence-key validation to Lemon Squeezy.