Modeling the Model Model

'Tis nobler to model?'

December 20, 2011

"To model, or not to model, that is the question:
Whether 'tis nobler to model or to suffer
The slings and arrows of a failed live pilot…"

The decision to create a simulation model of an improvement solution is a tough question. Not quite as tough as Shakespeare's soliloquy on the existential nature of life, but a tough decision all the same. There are many trade‐offs to consider, and the answers are not always clear.

The Goal of a simulation model is to provide greater insight into a proposed solution, insight that cannot otherwise be readily gained through other methods. So, the first question to ask is:

(1) "Is my proposed new process sufficiently complex that traditional LSS tools cannot sufficiently predict the performance and interactions that may occur in the process?"

Typical indicators that the answer to this question might be "yes" may include:

Task Variation—My process has to respond to a high degree of variation in the types of products or services it handles. High amounts of task variation in a process can often lead to unanticipated queues and shifting bottlenecks. A simulation model can help identify the potential range of queue lengths and utilization levels for the process.

Demand Variation — High variation in demand can also lead to unintended effects in a process. A traditional approach to ‘sizing' the capability of a process with high variation in demand is to perform steady‐state calculations at the low volume and high volume states. The results of those calculations then help us determine staffing levels, inventory levels, etc. for the process. However, the steady‐state calculations don't give us any information about the transitions between high and low demand points. How early should we add staff? How long should we retain staff after the high demand ‘rush' has past? Simulation modeling can help determine the performance of the process during these dynamic periods of time.

Process Variation — Think about a finance department. In fact, think of it as job shop where we have several similar (but not equal) resources and jobs (month‐end close, capital investment, cash flow management, etc.) flowing through the department. Just as in a manufacturing job shop, it is difficult to anticipate how a process will perform if the work flowing through the process can take many alternative paths in the process. Bottlenecks are constantly moving and queues can spontaneously form where you least expect them. Complex processes like this can benefit tremendously from a simulation model to help define the initial process itself, and to ‘stress test' the process in the future when new or special projects are being considered. The simulation model can help determine if the process can handle the changes.

The next question to ask is:

(2) What is the cost comparison between running one (or more) ‘Live' pilots versus building the simulation pilot?

The following table compares some of the risks and benefits of Live versus simulated pilots:




Cost to pilot

May be high or low depending on the process and the nature of the changes being made.  In general Live pilots tend to be expensive.

Creating a simulation model does take an investment of time, including the upfront investment of time to learn the simulation software itself.

Flexibility to try many ideas

Live pilots tend to have little flexibility, with each new idea requiring a new pilot (which requires more time and interruption to the process).

The main benefit of simulation is the ability to try many different improvement ideas in the safety of the virtual, simulated environment.

Risk of losing momentum

A failed pilot can put a big dent in the credibility of a change effort, and resetting to try a second pilot will take time. However, a successful first pilot can be a great way to build momentum in a change effort.

Building and validating a simulation model will take time.  If the likely improvement solution is apparent, taking time to simulation the solution may cause an unnecessary delay and loss of momentum.  However, avoiding a failed live pilot will almost always justify the time spent on simulation.

A step to far

Live pilots have the benefit of being live!  The pilot may fail, but you will have the benefit of collecting data and learning from the failure.

Extrapolating improvement ideas well beyond the underlying data and assumptions used to build the simulation model is always a risk.  Making too many changes and getting too far beyond the original process comes with the risk of your model assumptions breaking down.

And the final question:

(3) Do I build a simulation or not?

Maybe… Simulation is underused in today's LSS world, and part of that is due to the high investment required to learn most of the simulation software on the market. has created Process Playground to help address this gap. Process Playground is easy to use and easy to learn. Our software takes the 90/10 approach — we can do 90% of the models you may need to build, with 10% of the effort. After you have become a pro at building models with Process Playground, then you may move on to a larger simulation package when your models need to jump up to a high level of complexity.

Simulation can be a powerful addition to your toolkit for process improvement, and it can open up new, out‐of‐the‐box thinking about your processes. But it must be tempered by careful cost/benefit analysis to insure that the building of simulation models remains a means‐to‐an‐end, and not an end in itself."

Dr. Lars Maaseidvaag