Schedule risk and Monte Carlo simulation, explained

Why a single end-date is almost always wrong

Every task in a plan has an estimate, and every estimate is really a guess at the middle of a range. String dozens of these together and the odds that all of them land on or under their estimate are vanishingly small. This is why plans that "should" finish in June so often finish in September: the single-point date is the most optimistic plausible outcome, not the likely one. Risk only ever adds up in one direction.

What Monte Carlo simulation actually does

Instead of one number per task, you give each task a range — typically three points: optimistic, most likely, pessimistic. The simulation then:

1. picks a random duration for every task, drawn from its range;
2. runs the whole schedule once and records the finish date;
3. repeats this thousands of times.

The result isn't a date — it's a distribution of thousands of possible finish dates, each as likely as the inputs say it is. You're no longer asking "when will it finish?" but "what's the chance it finishes by X?"

Reading the result: P50, P80 and the S-curve

Sort every simulated finish date and you get a cumulative curve (the "S-curve"). The key readings:

• P50 — the date you'll hit or beat 50% of the time. A coin flip. This is usually close to the naive plan date.
• P80 — the date you'll hit or beat 80% of the time. This is the one most organisations commit to externally, because P50 fails half the time.
• The gap between them is your honest contingency. If P50 is March and P80 is May, two months of schedule risk are real and quantified — not a number someone padded by feel.

Critical-path risk and why the "critical path" can lie

In a static plan, one path is "critical" and the rest have float. But under simulation, a near-critical path with high uncertainty can become the path that actually drives your finish more often than the official critical one. This is criticality index — the percentage of simulation runs in which each task lands on the critical path. A task that's "not critical" but critical in 70% of runs is exactly where your attention belongs. Single-point plans never reveal this.

Garbage in, garbage out — the honest caveat

A simulation is only as good as its ranges. If every task gets a lazy ±10%, you'll get a confident, narrow, and wrong curve. Good schedule risk analysis spends its effort where it matters: realistic ranges on the tasks that actually carry uncertainty, and explicit risk events (a permit, a vendor, a dependency) modelled separately. The simulation does the arithmetic; your judgement supplies the risk.

The standards behind it

Quantitative schedule risk analysis is recognised practice in the PMBOK® Guide and the Project Management Institute's standard on risk; the three-point estimating it builds on is classic PERT. PM Vault aligns to these and cites them as authority — see our sources — and our simulation instruments publish their random seed, so the same inputs reproduce the same curve, every time.

You can run this yourself

Schedule-risk simulation used to mean a five-figure enterprise add-on. It doesn't have to. A focused instrument runs thousands of iterations on your own plan, on your machine, and shows you the P50/P80 curve and the criticality index — glass-box, offline, reproducible.