Simulation Methodology

We use simulation models wherever possible because of their power and ease of testing. They allow us to evaluate a wide range of action alternatives at low cost. They also help us understand more fully the nature of the process being studied, often illuminating important new process features and improvement opportunities.

The downside is that these models are usually difficult and time-consuming to build and validate. But once you have a model that works, tailoring it to new situations can be relatively quick and straightforward.

What is a Simulation Model?

For those who are not familiar with simulation models, here is a brief introduction. A simulation model is typically a special type of computer program that attempts to mathematically represent an actual system or set of activities in a realistic manner. It is designed to behave in a manner closely similar to the underlying real system.

Good simulation models, given actual system initial conditions, can generate output in the form of key system variables that are close to what the actual system would generate. This is in fact the primary way to validate most simulation models: adjust the model's structure, rules and process parameters until it can reproduce actual system outputs given actual system initial inputs.

Models are especially valuable for understanding and testing the impact of changes to complex systems such as business processes. Designing the model nearly always adds greatly to an understanding of the process. Validating the model by using it improves understanding of how the actual process behaves. If the model behaves very differently from the actual process, then the problem is usually a defect in the model's structure or rules. Proper behavior in general terms but not close enough in detail is usually the result of model parameters being incorrect.

Optimization

Some simulation models do not optimize but simply represent with acceptable accuracy an underlying process. Optimization, such as driving toward a minimum cost in the presence of various constraints, can usually be added to a simulation model. In many real-world situations, optimization is not possible because of the system complexity. Instead, we simply run the model over a wide range of conditions and select the "best" one, however we decide to define "best" in a given situation. 

Notes

Some folks make the distinction between a model, which is the computer representation, and simulation, which is use of the model under various input conditions. This is an academic distinction in our view.

Many business process simulation models are quite large computer programs attached to often quite large databases. Output is typically in the form of a standard office format such as available with Microsoft Excel, Access and PowerPoint.

Model design and application is as much art as science. You learn most of the important aspects by actually building and running models. This is how you find out what works and what does not.