Simulation and scenario intelligence.
A scenario is not a forecast, and not a promise.
What happens to cash if a supplier slips two weeks. What happens to the promise dates if the van breaks. Simulation answers those over your own ledger — and is careful to say that an answer is a scenario, not a prediction of what will occur.
Four ways of asking “what if”.
Deterministic scenarios
Change an input, re-run the same mechanics, compare. No randomness, and the same answer every time.
Monte Carlo
Sample from the distributions the predictive engine produced, many times, and read the spread rather than the middle.
Discrete-event simulation
Run the process forward event by event, with queues and capacity, to see where it congests.
Sensitivity and stress
Which input actually moves the answer, and what happens at the edge rather than at the average.
Because they answer different questions.
A forecast estimates what is likely. A simulation explores what would follow if something specific happened, including things that are unlikely and precisely for that reason worth testing. Conflating them is how a stress test ends up quoted as an expectation, so they are separate engines with separate outputs and separate labels.
- A simulation takes the distributions from predictive machine learning as inputs, rather than a single expected value that will not hold.
- It takes the mechanics from the deterministic engine, so the arithmetic inside the scenario is your real arithmetic.
- It takes process durations and capacity from process mining, so the queues behave the way your queues behave.
- It hands plan comparisons back to optimisation, so a plan can be chosen for how it behaves under stress rather than how it scores at the average.
A scenario is not a forecast, a commitment or an authorisation.
This boundary exists because scenario output is unusually easy to misuse. A stress case carries the label that it is a stress case; a bounded digital twin is bounded and says so; and no scenario output can authorise anything on its own. If a scenario suggests an action, that action goes through the same governed lifecycle as any other.
Before you commit, not after.
- “What happens to cash if this supplier slips two weeks?” — the exposure across every order standing behind it.
- “What if we take this rush order?” — what it displaces, and what that costs elsewhere.
- “What service level can we actually hold?” — a what-if over safety stock rather than a target somebody typed.
- “Which of these two plans survives a bad week?” — the comparison that the average-case score hides.
- “How sensitive is this to the lead-time assumption?” — the input that actually moves the answer.
Frequently asked questions.
Is a simulation result a prediction?
No, and the distinction matters. A prediction estimates what is likely to happen. A simulation explores what would follow given an assumption you chose — often deliberately an unlikely one. The output is labelled as a scenario for exactly this reason.
Where do the input distributions come from?
From the predictive engine, with their intervals intact. Using a single expected value instead is how simulations end up confidently wrong.
Can I run a scenario against live data?
Yes — it reads your live records. It writes nothing, and nothing it produces can authorise a change.
What is a bounded digital twin?
A model of part of your operation, explicitly scoped, with its limitations stated. The boundary is part of the output rather than a caveat somebody forgot to mention.
See this engine on your own records.
Join the waitlist and ask it something real. Every answer names the engines it used and the records they read.
Join waitlist