Six Sigma’s define, measure, analyze, improve and control methodology is well known as the process improvement program that “fixes” problems resulting from variability in manufacturing, engineering or transactional processes. There are times, however, when no fix will enable an existing process to meet customer expectations. A new process is needed to replace the old one, which leads to the question, “Can Six Sigma help design a new product or process?”
The answer is a resounding yes, with DMAIC’s companion methodology, known generically as Design for Six Sigma (DFSS). As with Six Sigma DMAIC, the DFSS methodology doesn’t actually design a new part or process since every enterprise has a unique design process tailored to its own product or service. However, DFSS can make these processes more robust and less costly. It delivers products and processes that perform at higher quality levels than otherwise possible. In short, while DMAIC can be loosely characterized as a find-and-fix methodology, DFSS could be thought of as a preventive one.
DFSS methodology can help a design team fully understand its customer’s requirements and predict if the design will meet those requirements at each phase of the process. It’s this predictive ability that saves costs in design time, prototypes and validation tests, which translates to a less expensive launch.
DFSS is defined as a systematic methodology, with tools, training and measurements that enable us to design products and/or processes that can be produced at the Six Sigma level. As such, DFSS users must begin by understanding their customers, who will have many expectations of a product or service. Not all of these requirements are equally important. The first task is to identify which of these will be the focus of the DFSS effort. This involves listening to the voice of the customer (VOC), prioritizing customer responses and, most important, identifying a measurable target and range for these requirements. Hitting these targets with the design and staying within an established range (i.e., minimum variability) will ensure that customer expectations are met and also serve as a measure of the design’s success.
The second important factor in DFSS is to understand the capability of the processes. To do this, we must answer this question for each key process: “How often will this process cause us to fail to meet customer requirements?” Comparing customer requirements and process capability enables us to predict the level at which we’ll be able to meet customer expectations.
As with Six Sigma’s DMAIC for process improvement, a key concept of DFSS is understanding the relationship of inputs to outputs, the Y = f(x) relationship. This is also known as a “transfer function” or “prediction equation.” Transfer functions can be determined by several different methods. For the simplest processes, data can be readily obtained from a process map or product drawings. In some cases, they may be described from principles inherent in the design’s physics, chemistry or geometry. In less obvious situations, we might be able to develop a model to describe the relationship between inputs and outputs using design of experiments. Designers have even conducted experiments on finite element models to obtain these relationships.
Regardless of how the transfer function is obtained, knowing it allows us to predict the quality level of any design before production. In other words, we’ll know in advance whether a design will meet customer expectations and what we might need to change.
Other useful information we can learn from the transfer function is exactly what effect each input factor is likely to have on an output. Fully understanding these relationships allows us to adjust a design to hit a target and to choose settings that reduce variability. This is the concept of robust design. In addition to designing for performance targets and cost, we can now design for a specific quality level and measure progress toward that goal throughout the process.
A manufacturer of portable power tools wishes to improve customer satisfaction with a certain tool by reducing the amount of noise generated during its operation. The designers have determined that a primary source of noise is bearing slap in the conversion of rotary to reciprocal motion. The design change is intended to reduce this clearance. When presented with this task, most designers intuitively observe that the clearance can be reduced by increasing the ball diameter, decreasing the angle of the raceway, increasing the width of the slider or some combination of these three. Without understanding the relationships embodied in the transfer function, however, these conclusions will lead to a less-than-optimal solution.
From the geometry shown on the drawings in the figure to the right, the designer was able to describe the clearance between the ball bearing and the slider dimensions in terms of the components’ dimensions. Using this approach along with DFSS tools, a solution was found that simultaneously achieved the desired target for the clearance and minimized the inherent variability from the manufacturing process.
That solution called for a reduced ball diameter (a), an increased raceway angle (a) and increased raceway width (b). The underlying reasons for this can be seen in the analytic description shown in the figure below. All four factors have an effect on the clearance, but two of them affect variability as well. The optimum design strategy used the “shrink” factors to reduce the variability and the “shift” factors to put the design on target. This case illustrates that design solutions to reduce variability often aren’t intuitive and are missed in the traditional design focus on achieving a single-point design.
There are some important prerequisites for successfully implementing DFSS design principles. The first and most important is stability in critical processes. Prediction is the essence of DFSS, and prediction relies on understanding process capability.
Second, DFSS is inherently a cross-functional activity. Process stability might be the domain of operations, but customer requirements must come from marketing and be communicated in quantitative and measurable terms. Achieving designs that are robust to the inherent variation of key processes requires optimizing design parameters through engineering. Finding those optimums is the business of design. None of these activities can be accomplished in the absence of the others.
Finally, it’s critical that DFSS is implemented in an environment of accountability. Clear targets for performance, quality, cost and delivery must be established at the outset and rewards (or lack thereof) for the design team should be based on measurable achievement at the end.
DFSS isn’t just for engineering or manufacturing processes. It’s equally useful in designing new transactional processes. DFSS for transactional processes is easier to apply and produces results more quickly than for many engineering or manufacturing applications. The method follows the same flow as that for manufacturing applications; however, each step can be easier to accomplish because the laws of physics or chemistry usually aren’t involved.
For example, a call center operation can be designed using DFSS as follows:
Understand VOC (e.g., use a Pareto chart to organize the types of inquiries, survey to identify the acceptable wait time, etc.)
Classify the inquiries into those that do and don’t require operators’ assistance
Identify critical functions that must be performed by the process
Draw a process map that brings the information to the customers accurately and efficiently to meet or exceed their targets.
Identify key steps that must be robust and design them accordingly
Predict the performance of the new process using existing process data, simulations or other techniques
Design a test to validate the new process and demonstrate that it meets customers’ expectations as predicted before fully implementing the solution
This example is highly simplified; however, the entire system can be designed and implemented within a few months with great improvement in speed, accuracy and customer satisfaction, as well as a reduction in labor costs.
Many successful DFSS implementations have begun with pilot projects. Training in the tools and disciplines of DFSS is best accomplished with cross-functional design teams focused on a single project. It’s best to begin using DFSS early in the design process to take full advantage of the methodology. Having the guidance of a DFSS practitioner who has completed projects is an enormous help.
Phong Vu is CEO of the Dr. Mikel J. Harry Six Sigma Management Institute (www.ss-mi.com), which has partnered with Arizona State University to provide Six Sigma Generation III and business leadership training for corporations globally.
Kempton Smith is a managing partner with Mosaica Partners LLC (www.mosaica.cc), which developed and taught the first DFSS course for DuPont and has since fostered Six Sigma implementations at numerous companies.