by Steve Fleming and E. Lowry Manson
Six Sigma has enjoyed
very positive press for several years now, primarily because of the financial successes of companies like GE that have embraced the technique. GE has saved billions of dollars by following Six
Sigma's define-measure-analyze-improve-control methodology. However, many of the company's process breakthroughs came as results of combining DMAIC with basic observation, mapping and "workout,"
a structured approach in which a Black Belt gathers process operators to analyze a process visually in order to remove inefficiencies, redundancies and unneeded steps. The method works
effectively with simple and easily visualized operations. But how can organizations analyze complex processes that aren't so easily visualized? As GE and other Six Sigma champions will tell you,
the answer is process simulation.
What is process simulation?
- New product planning process
- ECN process cycle time
- Inventory management
- Disposition of discrepant material
- Help desk/problem resolution
- Service engineer dispatching cycle
- Quote preparation process
- Order to remittance cycle time
- Accounts receivable management
- Customer billing process
- Hiring (approval to offer letter)
- Salary planning process
- Internal help desk
- Computer requisition/distribution
Just as a computer video poker program
imitates a real poker game, process simulation offers a computerized model of an actual process. The model is created by inputting into a simulation program all of the critical operating
parameters--such as process inputs, process flow and utilized resources--of the real process. With the addition of specific information about how these parameters vary, the resulting model
will essentially duplicate the real process.
A process model allows analysts to view a process operation from a desk or conference room, and the operation, which may normally
take days or weeks to complete, can be run in seconds. Such rapid, simulated operation makes it possible to collect large data samples that might be impossible to glean during normal
process operation. Specifically, with process simulation, you can collect and analyze data, run experiments and make changes without interrupting the flow of "mission critical"
operations. Once you've modified the process to meet the customer's needs and confirmed the results on the model, you can then make the necessary changes to the real process, confident
that your solutions will prove successful.
Most commercial simulation packages use an analysis method called "Monte Carlo." With this
method, each consecutive run of a simulation randomly uses a different combination of input values from preselected input parameters. By running the simulation through many cycles, it's
possible to collect significant amounts of data in minutes using a computer. This high-speed operation allows analysts to look at long-term data rather than merely the short-term data
available from a real process. Long-term data offer a more precise picture of how a performance will vary over time.
Other important simulation techniques include computer visualization and "action logic." Visualization capability allows operators to create high-powered,
live-animation presentations for management and process owners so they can see how information or parts flow through a process before and after changes are
made. Action logic allows users to select different paths and activities based on conditions that are programmed into the model.
Process simulation in a Six Sigma context
With any Six Sigma project, the first steps are to identify the process's
customers, understand what's important to them and discover how they'll measure success. These measurements are called "customer CTQs" (i.e., "critical
to quality" from the customer's perspective). Any simulation must measure performance against customer CTQs. If the simulation isn't capable of providing
this data, then determining their value becomes difficult and the particular Six Sigma project may as well be considered an exercise in futility.
Once the CTQs are known, a detailed understanding of the process to be simulated is required. Usually, a person familiar with simulation techniques will
interview process owners and operators so that a detailed process map can be built. The process map should include all process steps, decisions that are made,
rework that's performed and places where humans or machines add value or touch the process. You'll also want to know how long it takes to perform each
step on the process map, along with which resources are available for doing the step and where the outputs from the step lead. Don't worry if you can't obtain
exact timing data for each process step. With most modeling tools, it's possible to enter a variety of distributions for every resource, process step and input, which
allows you to make "best guess" estimates for building the model. Once the model is built, you can evaluate and confirm any assumptions that have been made during model building.
When to apply process simulation
As with all extremely powerful tools, process simulation must be used correctly
and only in appropriate situations. Process simulation can be a real aid in the following circumstances:
The process is very complex and difficult to visualize. With a simple process,
process mapping combined with brainstorming techniques will usually generate improvement ideas. With complex and hard-to-visualize processes, however, a
working model will help you understand where your efforts will provide the largest payback.
The process involves many decision points. The more decisions a process requires, the harder it is to visualize all of the possible paths that may be
The project goal is to optimize the use of resources for a process. Resource utilization is hard to visualize, especially when the same resources are used in a
variety of actions throughout a process or when one resource is used for some portions of the process and another resource is used later on.
The goal of the project is to establish optimum-lot ("kanban") sizing for a
manufacturing process. Many manufacturing organizations use demand-flow technology, which involves formulas to calculate kanban sizes. These work well with
simple manufacturing processes, but the formulas can be much more difficult to implement when several suppliers are linked in the chain. A process model allows
you to visualize the flow of material through the factory floor.
A Six Sigma case study using process simulation
In our 12 years of using Six Sigma to solve problems all over the world, we've yet to find a situation in which the methodology can't be applied. In some cases,
the most efficient tool for solving a problem is one that comes in an advanced toolbox. That's the case with process simulation.
One of today's most common customer support processes is the "help desk" or "call center" process. At some level, every manufacturing or service organization
includes a team responsible for answering and addressing customers' questions and complaints. Yet, customer satisfaction ratings of the help desk are generally
very low. Although this process is common, it's also difficult to analyze with conventional Six Sigma tools. The measure phase usually identifies "time to
resolve an issue" and "quality of the issue resolution" as the two CTQs. When these are measured, performance is generally less than one sigma (i.e., defects greater than 50 percent).
In a typical help desk improvement project, the analyze phase becomes an exercise in collecting existing data--if it exists at all--from call center logs and
automated recording systems. Once data are collected, a Black Belt will use regression analysis to understand the potential root causes driving poor
performance. During the improve phase, a recommendation will usually be made for additional staff at the help desk. The Black Belt won't do a design of
experiment for the project because there's no way to experiment with a process that touches so many customers. The real problem, of course, occurs during the
control phase, when staff is added and no change occurs in CTQ performance.
What happened? Well, help desks are much too complex to analyze using basic
statistical tools. Most help desks have two or three levels of support. When a call comes in, it slides into a queue. When a level 1 person is available, he or she takes
the call. If level 1 can't resolve the issue, the call is forwarded to level 2. If level 2 can't resolve the call, it's forwarded to engineering or a similar support group.
Between each of these levels, the call may end up languishing in several more queues, or the customer may be asked to wait for a call back.
A more effective way of improving the help desk process is building a process simulation as the first step in the analyze phase. Once the simulation is built, the
Black Belt can validate the model against the real process by collecting whatever data are available for model inputs, running the model and statistically matching the
results with data collected during the measure phase.
Once the model is validated, characterized analysis can begin on it. The analysis
can take a variety of paths, depending on the information provided by the simulation package. Most will offer operational output data for all the process
steps, resource utilization data and any additional variables tracked throughout the process. When the data are collected, it becomes a fairly straightforward task to
statistically analyze it, identify bottlenecks, develop proposed solutions and rerun the simulation. In addition, DOE can be used to help interrogate the model. After
you've optimized the model, you can also test it over a wide range of inputs and measure the performance against the CTQs.
Once the results have been confirmed on the model, you'll enter the control phase knowing that you've tested the solution and no customer will be adversely
affected--a great advantage over the traditional approach of using customers to test improvement concepts. We've used process simulation within the Six Sigma
methodology several times to improve help desk operations. In one case, we found that performance and customer satisfaction could be improved by reducing
the total staff and routing problems to the level 2 support team sooner based on the problem type.
Process simulation risks
As with most tools, there are always potential risks along with the benefits. Process simulation is no exception. Here are the most common:
. Once you've developed some expertise in using the tool, you'll suddenly find applications everywhere. Although there are
many appropriate situations for process simulation, it's not necessary--or even desirable--to model everything. Some discretion is helpful.
Overkill. It's best to simulate only those portions of the process that will bring
about improvement. As author and statistical theorist George Box observed, "All models are wrong; some models are useful." Concentrate, then, on those portions
of the process that require improvements as identified in your Six Sigma project. There's no need to develop the absolutely perfect model.
Overuse. Process simulation is just one of many important tools in your Six
Sigma toolbox. Methodology is the key that unlocks Six Sigma's potential, not the tools themselves. Follow the methodology and use the tools appropriately to help make decisions.
Summing up simulation
Although it's unnecessary for all Black Belts to be simulation experts, each should
understand where and when simulation tools are appropriate and include them in their toolboxes. It's also unnecessary to use process simulation on every
process-related project, though, with some, process simulation could mean the difference between success and failure. Process simulation can be a genuine aid
when the process is very complex or hard to visualize or includes multiple decisions points at which each decision could potentially change a project's outcome drastically.
About the authors
Steve Fleming is CEO of SigMax Solutions LLC. He has a bachelor's
degree in electrical engineering, a professional engineering license and 24 years of GE Medical Systems engineering and manufacturing experience. In
addition, he has more than six years of Six Sigma and DFSS experience, including four business startups. E-mail him at email@example.com .
E. Lowry Manson is chief technology officer at SigMax Solutions LLC. He has a doctorate in physics and spent 23 years in engineering and
manufacturing within several GE businesses, including super-abrasives, aerospace, medical systems and lighting. Lowry is a certified Black Belt and
Master Black Belt with more than six years experience delivering results on high-impact projects.
E-mail him at firstname.lastname@example.org .
Letters to the editor regarding this article can be sent to email@example.com .