The first step in building a business process simulation model is ensuring the overall flow is correct. Developing an operational flow first involves walking through the process with a process expert in a controlled environment. The article “The Process Modeling Handbook: How to Gather the Right Information in a Flash” covers data collection. Once you have gathered all this information, go back to your computer and use it to create a flow diagram, but with a twist. This flow diagram has to work. It’s essential to develop this flow-only model before adding any other details, as it will help you to identify any issues with the process and make any necessary adjustments before proceeding.
Animate the model
Once you have created a flow-only model, the next step is to animate the flow. Animation forces you to work out the flow details. Because everything has to work, there won’t be any “and then the magic happens” sections. Flaws become obvious. The ability to ask the right questions improve. Some simple tricks make the animation-only model more valuable – remember, these suggestions only apply to the animation-only model. First, change the move time on routes relative to their length. Routes default to one minute, so a path ten times longer could be set to ten minutes. Trust me; this simple time change will make the animation-only model easier to follow and understand. Second, put enough time between the arrivals so that only one entity at a time runs through the process. Use the Periodic arrival and set the Time Between Arrivals large enough to meet the need.
You might ask, “I’m building a business process simulation. Why would I create an animation-only model?” Good question! I’m glad you asked. Because building an animation-only model first will save you a ton of time. The flow is the foundation of the model. If you haven’t created the proper foundation, you will rework the model after investing significant time. In summary, get the foundation correct, and the remainder of the model building is simple.
Simplify and correct
After creating the animated model, call in the process experts and secure their approval. After reviewing the animated model with the process expert, you will likely identify some changes, additions, or simplifications that need to make to the model. Make these adjustments and then perform a second review with the expert to ensure that the model is accurate and complete. Finally, when the model is correct, print it and have the process experts sign it. The signature forces the process expert to remember their contribution in a future meeting and avoid their manager saying things such as “that’s not the process.”
Don’t forget to save the Animation-only model for future use, as it will be helpful in meetings and gaining support for participation.
Timing and capacities
Once you have finalized the model foundation, the next step is to enter the timing and appropriate capacities for each step in the process. This next step of entries includes input and output queues and any necessary model logic.
Use actual data, when possible, to create time distributions for processing steps. Stat::Fit is a powerful tool for changing raw data into distributions. When data is unavailable and estimating is required, use a triangular distribution as a starting point. You can ask three questions to uncover the range of work times:
- What is the least amount of time required to perform this operation?
- How long does it usually take to complete this operation?
- What is the longest time required to complete this operation? (Work content only)
A triangular distribution might look like T(3,6,8). See the graphical depiction below.
If the triangular distribution does not fit the requirement, there is a feature in Stat::Fit, called the Distribution Viewer to help represent times better. For example, if the times supplied to the three questions provided the answers:
- Least amount of time = 4 hours
- Most likely time = 9 hours
- Longest time 25 hours
- But experience has shown the times to be heavily skewed to the left.
You could select and modify a more appropriate distribution using the Distribution Viewer. First, find this feature from the Utilities menu by choosing Distribution Viewer. Next, select a distribution, then manipulate the parameters to more accurately estimate the times in your system. The distribution viewer provides an example below:
The next step is to add realistic arrivals to the model that reflect the actual arrival patterns of entities in the process. Most companies use one of two types of routings: arrivals — forming a repeating pattern and historical data read into the model. Both are fast and simple to recreate in ProcessModel. Within a couple of minutes, raw data converts to patterns or imports. Again, good arrivals will help to make the model more realistic and accurate.
The final step in building a business process simulation is to enter and assign all the resources needed for the process, including shifts, availability, and assignments. Unfortunately, allocating resources is often one of the most complex parts of building a model — which is why you shouldn’t attempt to add resources in the first stage of model building. That is to say, make the foundation secure before adding complex structures dependent on the foundation.
ProcessModel includes extensive resource assignment capabilities to capture the intricacies of employee skill levels, speed of completing tasks, error rates, and shift assignments.
Verify and Validate
Verify and validate the model to ensure that it is correct and complete. For example, these steps might include the following:
A. Check the model against the real-world process it is meant to represent. You have taken the word of the process experts; now you need to have eyes on the process. Look for discrepancies between the model and the actual process and make any necessary adjustments to the model to ensure that it is accurate.
B. Compare the results of the simulation model to actual data from the real-world process. Several key performance indicators may be available for comparison in an extensive process. For example, corrections might be as simple as adjusting a distribution to estimate times or quantities. In other instances, the output statistics help you to uncover information the experts want to hide (because statistics don’t lie).
C. Another way to validate and verify a simulation model is to review it with additional experts familiar with the modeled process. They can help to identify any errors or inconsistencies in the model and suggest ways to improve it.
D. Test the model by running multiple simulations with different scenarios and inputs. Stress testing can help ensure that the model can accurately predict the outcomes of different situations.
By following these steps, you can build a simulation model more efficiently and effectively, minimizing the risk of backtracking and rework of the model.
Experience building a model for yourself with ProcessModel. Click Here for a free trial.
Would you like help in building a simulation model of your process? Click Here to request a call from a representative.
Read the article on how to collect data to build a process model.