When Should You Actually Use Simulation?

Published on January 14, 2026

Simulation is powerful. But it's not always the right tool.

Here's how to know when simulation earns its place—and when you're overcomplicating things.

Use Simulation When...

You Have Queues and Variability

If entities compete for limited resources and arrival or service times vary, simulation shines.

The spreadsheet shows you averages. Simulation shows you reality—the peaks, the queues, the frustration.

You Need to Test "What If"

What if we add another machine? What if demand doubles? What if we change the process?

Running experiments on a real system is expensive, slow, and sometimes impossible. Simulation lets you test dozens of scenarios before committing.

Interactions Are Complex

When Step A affects Step B, which affects Step C, which loops back to Step A—simple calculations fail. Simulation handles feedback, dependencies, and emergent behaviour.

You Care About Variation, Not Just Averages

Average waiting time: 5 minutes. Sounds fine. 90th percentile: 25 minutes. Not fine at all.

If distribution matters, if worst-case matters, if "most of the time" isn't good enough—you need simulation.

The Stakes Are High

Building a new hospital wing? Designing a production line? Planning an airport terminal?

The cost of getting it wrong dwarfs the cost of simulation.

Don't Use Simulation When...

Simple Maths Will Do

If your question is "How many hours to process 100 items at 10 minutes each?"—that's arithmetic, not simulation.

Don't build a model when a calculator suffices.

Data Doesn't Exist

Simulation needs input data: arrival rates, service times, failure probabilities. If you're guessing all the inputs, you're simulating your imagination.

Bad data in, bad insights out.

The Decision Is Already Made

Sometimes simulation is commissioned to justify a decision already taken. That's not analysis—that's theatre.

If the answer is predetermined, save everyone's time.

The System Is Stable and Simple

A conveyor belt running at constant speed with deterministic inputs doesn't need simulation. Simple queuing theory might be enough.

Simulation adds value when things are messy. Predictable systems need simpler tools.

You Need Real-Time Response

Simulation runs many scenarios to understand the distribution of outcomes. If you need instant answers in production, you might need a different approach (heuristics, optimisation, machine learning).

The Decision Framework

Ask yourself:

1. Is there queuing? Resources shared, entities waiting?
2. Is there variability? Randomness in arrivals, service, failures?
3. Are there interactions? Dependencies between processes?
4. Do I need confidence in the tails? Worst-case scenarios matter?
5. Is experimentation costly? Can't easily test on real system?

If you answered "yes" to multiple questions, simulation is probably worth it.

Simulation vs Alternatives

The Real Test

Here's the practical test: would a decision-maker change their choice based on simulation results?

If the simulation reveals that a proposed design has a 40% chance of failing targets—that changes decisions.

If the simulation confirms what everyone already knows—was it worth building?

Common Mistakes

Over-simulating Not every problem needs a 10,000-entity model with 50 parameters. Start simple. Add complexity only when it changes insights.

Under-validating A simulation is only useful if it reflects reality. Validate against historical data before trusting future predictions.

Model worship The model is not the system. It's a simplification. Every model is wrong; some are useful. Treat outputs as informed estimates, not gospel.

My Rule of Thumb

When someone asks "Should I simulate this?"—I ask them to describe the problem in one sentence.

If that sentence involves queues, variability, and a decision with consequences, simulation is probably the right tool.

If it doesn't, there might be a simpler way.

Get Started

Ready to try simulation?

• What is Discrete Event Simulation?
• What is SimPy?
• Build Your First Model