Skip to content

Monte Carlo

Re-run the critical path thousands of times over three-point durations and register risks, then commit a chosen confidence level back into the schedule as a real, rescheduled plan.

Who it is for. Anyone who has to state a P80 date or a P80 cost to a board, a lender or a client — and defend how it was derived.

What it is for

A deterministic finish date is a single sample from a distribution and everyone knows it, but quantifying that is normally a separate tool and a separate export. This runs the product’s own CPM engine every iteration, so a delay only moves the finish when it lands on the driving path.

Against the incumbents

Schedule risk analysis is a separately licensed product in the incumbent world. Two specifics go beyond "we have Monte Carlo too": running the same CPM engine as the live schedule, and applying a P-level result back as per-activity durations that reschedule into a valid plan.

Monte Carlo
P10, P50 and P80 against the deterministic finish.

What it does

Every line below is in the product today.

  • The real scheduler, every iteration

    Each pass samples every duration and re-runs the same CPM engine, link manager and calendars the live schedule uses — so simulated and deterministic behaviour cannot diverge.

  • Apply a P-level back to the plan

    Applying a confidence level writes each activity’s own sampled duration and reschedules, producing an internally consistent plan. Writing per-activity P-level dates would not, because they are correlated — and the code says so.

  • Register risks in the simulation

    Risks are sampled alongside duration uncertainty and injected into the activity they are linked to; opportunities apply the same magnitude with the sign flipped.

  • Activities that might not happen

    An activity sampled out is held in the network at zero duration, so its predecessors and successors still connect through it.

  • Criticality index

    The share of iterations in which each activity sat on the critical path, with iterations it was sampled out of correctly excluded.

  • Unquantified risks are counted, not zeroed

    A risk with no impact for the chosen metric is skipped and reported as skipped, so it cannot silently understate exposure.

  • One undo for the whole what-if

    Applying a scenario is wrapped in a single undo batch, so Ctrl+Z reverses the experiment rather than one activity of it.

  • Runs off the main thread, and cancels

    Simulations run in a Web Worker with live progress and a cancel that works.

See it on your own schedule.

Send an XER before the call and we will import it and show you Monte Carlo running against your data.