Why We Should Reposition Simulation in BPM

Often BPM implementations can be characterized by a struggle between process design and automation. Process design focuses on getting the process right before automation, while many BPM vendors consider it a mistake to spend too much time on process analysis at the expense of delaying automation. Traditionally simulation has been a tool for process design due to its predictive ability, and an assumption that direct experimentation with the business process under study would be disruptive or costly. The usefulness of simulation has been generally accepted to the point that it is now considered a key component of BPM suites. Yet despite this, simulation is a feature that is not widely used in BPM. I’d like to explain why this might be so, and suggest that we need to rethink how simulation is used in BPM to better leverage its enviably high predictive ability.



Figure 1: Simulation as a process design tool.

The Costs of Simulation

Testing process designs prior to automation appeals to our intuition. Better to build a highway than “pave cows paths” as the saying goes. However, making a business case requires that we examine costs versus benefits and place aesthetics and intuition slightly to the side. Under this light, simulation as a process design exercise makes the most sense when the costs of automation are relatively high when compared to the costs of simulation analysis. I would argue that historically this has usually been the case. Specifically, simulation itself has its roots in manufacturing where the costs of retooling or reorganizing a plant would be daunting without clear proof of the ultimate savings. However, process analysis and design does not come free. In the case of simulation, three costs in particular should be considered:

  1. Skill Set: As a process design tool, the creation of meaningful and useful simulation models requires specific skills. Proficiency with simulation modeling tools are typically components of an industrial engineering or management science curriculum yet simulation analysis is often performed by business analysts when used in the scope of business process management implementations. Without proper training, processes are often modeled incorrectly or not enough data is collected to properly parameterize the model.
  2. Data Acquisition: Simulation models require more than a process description. After the work flow has been modeled, simulation models must be augmented by a significant data set that describes such things as the volume and arrival pattern of work, task durations, resource levels and availability, and more. Typically historical sources for this data must be found, and/or forecasts generated. This data will need to be massaged and reformatted for use in the model, and great care taken to fill in the inevitable gaps. Furthermore, this exercise may need to be partially repeated if you are going to attempt to re use the model in the future. Data acquisition and preparation is perhaps the most significant and costly component of simulation analysis.
  3. Deferment of Automation: Process design and analysis will delay automation. The tendency to gold plate a process: the desire to make it perfect can lead to analysis paralysis. This represents an opportunity cost where any benefit based on automation alone will be delayed. Critically speaking, when a business case can be predicated on automation alone, one has to consider why a process design phase and/or simulation analysis is even warranted.

If the costs of simulation described above are significant, let alone anywhere near the costs of automation, the entire premise of this kind of process design and analysis needs to be critically reviewed.

Rethinking Simulation

The good news is that the costs of simulation come down significantly when used for analysis after automation:

Figure 2: Using simulationafter automation.

Obviously, by positioning simulation for use after automation, simulation analysis does not delay the benefits of automation. Automation also provides the kinds of data required for parameterizing simulation models virtually for free. Lastly, because of the easy availability of high quality, accurate data, and the potential for process discovery, the skills required to create meaningful simulation models is reduced as the simulation model construction process is also at least partially automated as a consequence of process automation. The predictive ability of simulation models also increases under these consequences because of the direct use of real process data and less reliance on estimation and probability distributions.


Figure 3: Costs and accuracy of simulation after automation.

Using simulation post automation requires that we reconsider simulation as not a process design tool, but instead a tool for real time, or near real time operational decision making. It is easy to see how simulation can be used in this context to provide both predictive and prescriptive abilities to management. Consider the following use cases:

  • Accurately predict completion times given the current state of the process and items currently being worked on.
  • Take the current state of a business process and anticipated volumes to generate accurate predictions of near term work load by skill set. Workforce management solutions can use this data to effectively and optimally schedule staff to ensure adequate capacity is available, while taking into account any staffing constraints such as the specific availability of specific resources.
  • Test changes to the process structure itself based on the real time circumstances and suggest optimizations to management, for example, the relaxation of certain business rules at a decision point in the work flow for a limited time frame.
  • Dynamic activity based costing where simulation is used to accurately predict the costs of processing individual items based on real time data, potentially offering unique opportunities for more effective pricing where appropriate.

It’s time for simulation technologies to move beyond being a tool for analysts and to start being used in production on an ongoing basis to help managers better manage their processes.

About the Author

John Januszczak designs solutions that optimize process, staffing and resource levels for organizations with complex processes. John has been engaged in simulation, software development and business process transformation for 20 years. John has worked in a broad spectrum of industries from telecommunications to financial services. Currently, John is working on the integration of process modeling and simulation technology with workforce management applications, as well as developing process analytics and simulation software as a service solutions. He is the founder of the Sim4BPM initiative. John is currently employed by Meta Software Corporation. He blogs at http://sim4bpm.com and can be reached at http://twitter.com/jjanuszczak.