Simulation as Experimentation

Definition

Simulation builds a model that relates controllable inputs (decisions) and probabilistic inputs (uncertain quantities) to an output of interest, then generates many randomly sampled values for the probabilistic inputs to see the resulting distribution of outputs — most often to assess risk of loss before committing real resources. The book is explicit about a boundary condition that separates it from optimization: "Simulation is not an optimization technique. It is a method that can be used to describe or predict how a system will operate given certain choices for the controllable inputs and randomly generated values for the probabilistic inputs."

In the Book

Chapter 12 introduces simulation through PortaCom's decision to launch a new portable printer, where selling price ($249) and marketing costs are known but direct labor cost, parts cost, and first-year demand are uncertain. The book first tries what-if analysis by hand: plugging in a base-case scenario (profit of $710,000), then a worst-case scenario (all three uncertain inputs at their worst simultaneously, a $847,000 loss) and a best-case scenario (a $2,591,000 profit) — and notes this leaves management with only three points, no sense of how likely each is. The chapter then generalizes the approach to a full simulation that samples probability distributions for each uncertain input many times over, producing a distribution of possible profit outcomes rather than three anecdotes. The same structure is reapplied later in the chapter to an inventory-policy problem and to simulating the Hammondsport Savings Bank ATM waiting line, and the "Call Center Design" case shows a real company using waiting-line simulation to size a call center's staffing before changing its actual service program.

Why It Matters

Simulation lets a decision-maker rehearse a system's behavior under uncertainty — including its downside risk — without experimenting on the real system, and without pretending a single best-guess forecast captures what could happen. It sits deliberately apart from optimization: rather than solving for a single "best" answer, it shows the full range of outcomes a decision could produce, which is often the more honest and more useful question when the future is genuinely uncertain rather than merely unknown.