User's guide

Scillion is a causal dynamics engine for modeling how systems evolve over time. You define parameters — either stateful (their value carries forward and changes each step) or stateless (their value is recomputed fresh each step) — wire them together with operations, and run simulations to see what happens.

The core idea is simple: most things you care about in a business or system don't change in isolation. Revenue depends on customers, customers depend on acquisition and churn, churn depends on product quality. Scillion lets you make those dependencies explicit and simulate them forward.

Models are built from equations. Every relationship is visible and editable — there's no black box. The engine is differentiable, so after a simulation it can tell you which inputs had the most influence on a given outcome, computed in a single backward pass.

This works best when you have domain knowledge but limited data. If you understand your system well enough to write down the rules, you can simulate it. You can also fit parameters to historical time-series, or use real data to validate model behaviour.

It's not a forecasting tool in the traditional sense — it's a structured way to think through how your system works, make your assumptions explicit, and stress-test them.