Building a Simulation

Building a simulation starts with a process or value stream map of your current process. This map will describe the basic flow of items through the various blocks, each of which represents a step or component of the process. Notably, these maps often do not include the detailed data needed to build a simulation model. For that, you’ll need to quantify precisely how each item interacts with each component of the model.

Here's how to build a simulation model:

1. Map your process: Identify all the key steps in the process, along with the different paths items take. In building a simulation model, it is important to consider the scope of the model. Model too much of the process and you may spend all your time on a portion devoid of problems or unable to be changed.

2. Determine the items that are flowing in the process: This step can be simple or complicated, depending on the process. For many processes, you can treat your entire population of demand as a single group or item. For some, you may need to spend time segmenting your demand into different groups that have unique flow paths and impact blocks in your model differently.

3. Determine the demand arrival and processing time distributions: In this step, you collect data and analyze it. In this analysis, you will determine the best mathematical distribution to characterize how items arrive to your process and how they interact with each block in the process. You will find statistical tools helpful in characterizing the data and identifying the appropriate distributions to use in the model.

4. Collect descriptive statistics of your live process: Understand the current performance of your process by collecting output statistics such as inventory levels, utilization percentages, and throughput counts.

5. Validate your simulated process: Run your simulation model and compare the output statistics of the model to those of the live process. If the statistics don't match, you'll need to investigate and continue to refine your model until it matches reality.

At this point, you will have created a validated model that accurately represents your true process. From here, you can begin experimenting with the simulated process to pilot various changes and improvements to understand their effects. If you have a capacity problem, you could experiment with adding more people to a process or improving workflow while keeping headcount flat.

The possibilities are endless - and limited only by your imagination. Keep in mind that the more your experiments diverge from the existing process, the less accurate the simulation will be in predicting performance. Start with small changes and understand their effects, then move on to larger, more dramatic changes. With these changes, think about how you can test your assumptions about them through small-scale live pilots. Here, you can re-verify the changes before committing to a large-scale process overhaul.