Move the levers. Watch the call change.

A toy model, stated out loud.

A calculator that hides its formula is a magic trick. Here is the whole thing — small enough to disagree with, which is the only reason to trust a number at all.

The recurrence

Start with an empty book. Each month, keep (1 − churn) of it and add the new customers. Run that twelve times, multiply by price. That is the whole model.

Downside and upside

Same arithmetic, churn moved by +3 and −2 points. Churn is the lever with the most convexity over twelve months, so it is the one worth bracketing.

What it ignores

Expansion revenue, seasonality, ramp time, discounting, annual prepay, and the fact that raising price usually moves churn. A real forecast has to name those.

Where it runs

In your browser. No account, no server call, no model. Nothing you type here is sent anywhere or stored.

What a real one adds.

Inside SynthBoard the same shape is authored by a named expert from a bench of 24, and it carries the three things this page cannot:

• Your numbers, not sliders.

  It reads what you have already told your board — price, churn, runway — and says where each figure came from and when you last confirmed it.

• The assumption it is least sure about.

  A forecast that will not name its weakest input is a guess wearing a chart. Every assumption comes back as a labelled input carrying its own number — marked as your figure or the expert’s estimate — and each one declares how far the answer moves when you change it.

• What would change the call.

  The specific observation that should make you redo it — so the number has an expiry date instead of a vibe.

Ask your own question free and the router seats whoever should answer it — or read how we hold your data before you tell it anything.