Designing for Six Sigma Capability

This article is the first of a two part series. The second part will appear in the March issue and will address optimizing existing processes to achieve 6 capability. The ultimate in optimization is called Six Sigma (6σ), and its practitioners are known as Black Belts. It may sound like a mystical art, but it is in fact a data driven method for achieving near-perfect quality.

By Dave Harrold, CONTROL ENGINEERING January 1, 1999
  • Process control & instrumentation

  • Productivity, measurement & control

  • Quality assurance

  • Statistical quality control

Is 99% (4σ) good enough?
Is Six Sigma software good enough?
CompactPCI systems approach Six Sigma capability

This article is the first of a two part series. The second part will appear in the March issue and will address optimizing existing processes to achieve 6 capability . The ultimate in optimization is called Six Sigma (6σ), and its practitioners are known as Black Belts. It may sound like a mystical art, but it is in fact a data driven method for achieving near-perfect quality.

In the 1980s a customer satisfaction ‘revolution’ swept across U.S. businesses promoting the concept that focusing on customer satisfaction was good for business. Books with titles such as Customers for Life , by Carl Sewell & Paul Brown, Quality is Free , and Quality Without Tears , both by Phillip Crosby, were being read by many business leaders. Employees were attending quality improvement courses, forming quality circles, and organizing customer satisfaction improvement teams. For some companies, investments never produced the desired results. In fact, Tom Brokaw, NBC News, recently reported that many companies embracing the customer satisfaction philosophy in the ’80s have returned to a profit driven philosophy for the ’90s. For others, such as Motorola, Inc. (Schaumburg, Ill.), General Electric Company (GE) (Fairfield, Conn.), and AlliedSignal Inc. (Morristown, N.J.), achieving customer satisfaction means good business, and good business equates to good profits.

Patience pays big dividends

Providing quality products and services is at the heart of customer satisfaction. For those companies who stayed the customer satisfaction course, financial gains have been good. For example, between 1987 and 1997 Motorola’s sales grew at a compounded growth rate of 17% per year ($27.79 billion in 1997); profits compounded at 17.2% per year ($1.18 billion in 1997); and stock value enjoyed a double-digit growth of 16.5% per year. Additionally Motorola’s cost of poor quality per unit has reduced more than 84%; employee productivity has increased 204% (a 12.3% per year average); and product reliability has improved better than five fold.

AlliedSignal added 6σ to its Total Quality process in 1994. Raymond C. Stark, AlliedSignal vp of Six Sigma & productivity explains, ‘Our quality process provides a foundation for continuous improvement. Six Sigma added the ability to achieve major change strategies throughout the company. Overall applicationof 6σ principles saved AlliedSignal $500 million in 1998, $1.5 billion to date, and is projected to save an additional $600 million in 1999.’

GE launched a corporate wide quality improvement strategy in 1995 when Jack Welch, chairman and ceo committed GE’s empire to reach 6σ quality by year 2000. Operating at about the industry average of 3 sigma in 1995, Mr. Welch’s vision is a stretch goal, but GE is making progress to achieve 6σ performance and estimates that 6σ will contribute an extra $5 billion to net earnings through the end of the century. In 1998 Mr. Welchexpects 6σ to yield about $750 million in net benefits.

Using statistical techniques to produce quality products is not a new concept. Pioneers such as Walter Shewhart, W. Edwards Demming, and Joseph Juran brought quality to the forefront. Following WWII, quality was embraced by many Japanese companies. Thirty years later Phillip Crosby introduced the concept of zero defects. But it wasn’t until the ’80s, when Motorola institutionalized 6σ to improve production of pagers, cellular phones, and other products, that statistical process control caught the eye of business leaders. Motorola’s success with 6σ techniques popularized it as a tool for product improvement. Today 6σ techniques have been expanded for use in improving every facet of business, from marketing, to order entry, to technical support.

Sigma Significance

Sigma numbers Defects per million
1.5σ 500,000
2.0σ 308,300
2.5σ 158,650
3.0σ 67,000
3.5σ 22,700
4.0σ 6,220
4.5σ 1,350
5.0σ 233
5.5σ 32
6.0σ 3.4
Six Sigma isn’t twice as good as three Sigma, it’s almost 20,000 times better.
Source: Control Engineering

Historical process capability

Until recently, a process was judged to be satisfactory with a 3σ capability. This means that if process control limits were placed on a process capability curve, the upper control limit (UCL) would be at 3σ to the right of center and the lower control limit (LCL) would be 3σ to the left of center (see Sigma capability curve). The area under the curve betweenthe two control limits (99.73% of the total area) represents the products or services conforming to specifications. The area outside the control limits (only 0.27% of the total area) represents an out-of-spec product or service. When converted to defects per million (DPM), 0.27% equates to 2,700 DPM. Statisticians have found processes often shift up to 1.5σ from center. When a 3σ process shifts 1.5σ from center, only 93.32% of the area under the curve remains inside the control limits. This equates to 67,000 DPM. When a process obtains 6σ capability and the same 1.5σ shift from center occurs, the process produces only 3.4 DPM (see Sigma capability chart).

Achieving total customer satisfaction requires a complete process and obtaining 6σ capability is proof the process is working. At Motorola it’s called Quality Systems Review; AlliedSignal and GE simply call it Six Sigma. Regardless of the name, these are well-developed processes tuned to produce excellent results.

In its purest form, 6σ is a measurement and analysis tool, but knowledgeable practitioners know quality can be designed in. Using a structured approach, a robust design can be developed. Robustness is quantified by a capability index (Cp) which is the ratio of the maximum allowable range of a characteristic to the normal ±3σ variation. For example, a 6σ design will yield a Cp equal to 2 ((UCL – LCL) / 6σ = 2). Designs having a Cp of 2 or greater are capable of producing extremely reliable products or services.

GE Plastics (Mt. Vernon, Ind.) uses a formal method to design robust chemical processes to produce engineered plastics such as LexanT, UltemT, ValoxT, and NorylT (see Designing for Six Sigma capability chart). Using a series of ‘tollgates’ GE Plastics’ Design for Six Sigma (DFSS) methodology guides project development and insures best practices are consistently used. Within each tollgate detailed key elements must be completed before the project progresses. Each element is supported by detail documents, spreadsheets, and databases. As information is entered into the DFSS software, completed key element boxes change colors. Approval to advance past a tollgate cannot be obtained without confirmation by both the software and business leaders. Following a best practices design methodology permits early identification of most project risk. Knowing the risk, designers eliminate or minimize risk so project goals and expectations are met. While following a formalized design methodology is not new, discipline to follow it has been sporadic in many companies.

Key elements appearing in best practices design methodologies include:

  • Understanding critical to quality (CTQ) external customer requirements;

  • Understanding CTQ internal customer requirements;

  • Conducting failure mode and effects analysis (FMEA);

  • Performing Design of Experiments (DoE) to identify critical variables; and

  • Benchmarking to remove ambiguity.

Identifying, documenting, and ensuring external and internal customer requirements are understood and built into the solution, whether for a product or a service, is critical. Bill Stackhouse, GE Fanuc (Charlottesville, Va.) manager of hardware development says, ‘The Six Sigma process helps us define what our customers consider ‘critical to quality’, or what we call CTQs.’ An example of meeting external and internal CTQs in a product design are apparent in the labeling of GE Fanuc’s VersaMax PLC and I/O systems. VersaMax is manufactured for distribution by GE Fanuc and OEM (original equipment manufacturer) customers. The OEMs want to specify the plastic case colors (an external CTQ). Producing an environmentally friendly product that was easy to inventory, and easy to recycle was an internal CTQ. To meet these CTQs, GE Fanuc needed to replace paper labeling with laser etching. To make the etched information clearly visible in different case colors, GE Fanuc worked with their plastics supplier (GE Plastics) to engineer a plastic that provided contrast for easy readability after laser etching.

An effective process control system requires careful research in the planning phases. Potential failure modes of each process, product, or service must be explored and understood. An initial FMEA must be conducted to assess potential problems. Knowledge gained during this investigation is used to plan the control system. Information-such as determining which important characteristics to be controlled, how to control them, how to react when the process becomes unstable, etc.-is necessary in maintaining processes within the upper and lower control limits.

In a complex process, with multiple interactive variables, experiments are necessary to determine the influence variables have on process output. Performing focused experiments isolates the ‘vital few’ variables from the ‘trivial many.’

Two popular DoE methods used by 6σ practitioners are classical and Taguchi. Carol Kavanaugh, C.F. Kavanaugh & Associates (Kingston, Ontario Canada) explains, ‘The major difference between classical and Taguchi DoE methods is the first does not require the experimenter to make assumptions about the presence or absence ofinteractions before the experiment runs.’ Regardless of the DoE method selected, designing a product or service to provide 6σ capability cannot be completed until the critical variables affecting output are identified. (For additional information comparing classical and Taguchi DoE methods, visit .) In the past, many projects were approved and implemented based on ‘gut feeling’ and intuition. Designing a process to deliver 6σ capabilities requires quantitative data. Using ‘gut feeling’ in a Design for Six Sigma methodology will not permit passing a ‘tollgate.’

Benchmark data can be acquired from existing similar facilities, pilot plants, competitive analysis, purchased databases, surveys, and Internet research.

Sample benchmarks

IRS phone-in tax advise (2.2σ)
Restaurant bills, doctors prescription writing, and payroll processing (2.9σ)
Average company (3.0σ)
Airline baggage handling (3.2σ)
Best in class companies (5.7σ)
U.S. Navy aircraft accidents (5.7σ)
Watch off by 2 seconds in 31 years (6σ)
Airline industry fatality rate (6.2σ)
Source: Control Engineering, Motorola, and GE

Process step A produces poor yield, step B has the highest cost of poor quality,
and step C has the worst capacity-productivity. If these three projects yield
only half the benefits, the process and profits are greatly improved.

Selecting projects, people, tools

Six Sigma methodology is a simple concept. Identify the right breakthrough projects. Identify black belt candidates best qualified to work on the breakthrough projects, and for four months, intermix formalized training and full-time focus on using the right tools and methods to successfully complete the projects.

To get 6σ off and running, projects and people are selected prior to training on how to use the tools. Dr. William J. Hill, AlliedSignal research fellow and master black belt, explains, ‘The price of admission into our 6σ training is a qualified project. This avoids training for the sake of training.’

Determining qualified projects requires understanding company metrics, such as rolled throughput yield, cost of poor quality, and capacity productivity (see Selecting the right projects).

Rolled throughput yield (Rty) is the product of the first-pass yield at each major process step. If a process has a first-pass yield of 90% at each of three steps, the Rty is calculated as 0.9 x 0.9 x 0.9 = 0.73. This result indicates that 73% of the first-pass product produced through the process is within specification.

Cost of poor quality (COPQ) includes costs of failing to produce and deliver 100% quality to customers. These costs include costs of all activities due to a product that is not first-pass quality such as rework, sorting, and blending. External COPQ can also include warranty costs and repair of customer returns.

Capacity-productivity (C-P) tracks the progress of increasing throughput, or reducing cycle times of manufacturing operations or services.

Kenneth McIntyre, AlliedSignal technical project specialist, has been instrumental in refining AlliedSignal’s activity based management (ABM) efforts and says, ‘Moving from direct costing to ABM has been very beneficial in helping 6σ to be successful. Six Sigma tools rely on development of process maps and ABM provides a process view of the business, making it easier to assign dollar values.’

One reason many customer satisfaction or quality efforts fail to meet expectations is because they are added to employee duties. At Motorola, AlliedSignal, and GE, full-time teams manage, promote, conduct training, improve the quality process, and share best practices. Because permanently removing ‘evil’ product and service defects requires diligence, special training, and unique skills, Six Sigma practitioners proudly carry the titles Master Black Belt, Black Belt, and Green Belt. (A few 6σ programs include levels such as Yellow Belts, White Belts, etc.)

Similar to progressing through a martial arts discipline, 6σ encourages people to continually ‘hone’ skills and work to achieve higher degrees of proficiency.

People usually learn faster and retain more of what they learned when they can immediately apply their new knowledge. Because Black Belt candidates are carefully matched to qualified projects, enthusiasm and success rates are very high. Dr. Stephen Zinkgraf, senior vp Sigma Breakthrough Technologies (San Marcos, Tex.), and a long-time evangelist of 6σ says, ‘Black Belt’s in chemical plants are averaging $300,000 to $500,000 savings on their first pass through a process area. What chemical operations learn from these early projects are the ‘knobs’ and the impact each ‘knob’ has on performance, throughput, and quality.’

Qualifications of Master Black Belts vary by company, but first and foremost, Master Black Belts are trained teachers, mentors, and deliverers of project savings. Master Black Belts meet all qualifications of a Black Belt, and have had direct responsibilities for at least 20 successful breakthrough projects. AlliedSignal Master Black Belts also have at least one year of advanced technical and business training. Final approval to reach Master Black Belt status comes from a committee of business leaders and peers.

Black Belts usually have college degrees and are working in their field. Black Belts receive four weeks of training spread over four months and use a preselected project as the basis for the training. Following training, Black Belts provide full-time high-level breakthrough project leadership.

Green Belts often have college degrees and are working in their field. Green Belts work under the watchful eye of a Black Belt and lead smaller projects. Green Belts receive 3 to 6 days of 6σ methods training, spread over 2 to 4 months. Training is conducted on preselected projects. An example of a Green Belt project assignment is a need to improve a measurement system that is part of a larger project.

Japanese Samurai carried seven tools into battle. While the tool set of 6σ methods is expanding, there are seven basic tools each 6σ practitioner is proficient and carries into the ‘battle’ to improve productivity and quality.

  • Histograms often appear as bell curves and show mean range and distribution;

  • Cause-and-effect diagrams provide visual presentation of conditions that effect outputs. For example, control system interlock screens;

  • Check sheets;

  • Pareto diagrams show in descending order what factors contribute the most defects;

  • Graphs;

  • Control charts plot sampled data representing process output with respect to time in a form that makes it easy to determine if the process is in statistical control; and

  • Scatter diagrams can show stratification when data is correlated. For example, raw material data correlated to different suppliers can produce meaningful information.

The ability to use the basic tool set is a prerequisite to 6σ Black Belt training where emphasis on using advanced quality and statistical skills is focused on four major phases of process improvement; Measurement, Analysis, Improvement, and Control.

Experienced 6σ practitioners recognize first-pass projects reap benefits almost as easily as picking up fruit from the ground. Subsequent projects in the same area require better use of tools, better data interpretation, more creative ways to solve problems, and introduction of newer, more sophisticated tool sets. Dr. Hill explains, ‘Tool set changes are occurring in two areas. First, idea generation tools incorporating quantitative rigor and using statistical basis can capture market needs, and critical customer requirements to assist in development of new ideas. Second, as we revisit processes to harvest higher hanging fruit, simulations and models are needed to assist in understanding fundamental relationships and processes to increase yields and reduce defects. Also, conducting DoEs on models removes risk of upsetting the actual process.’

Several pioneering companies such as Motorola, AlliedSignal, and GE have fine tuned the Six Sigma Design process to gain competitive advantages. Fortunately, companies who choose to adopt Six Sigma Design methods today can benefit from the efforts of these early pioneers. Going to and typing in Six Sigma Design will locate more resources and information about 6σ than you can imagine. So what are you waiting for?

Is 99% (4σ) good enough?

20,000 lost pieces of mail every hour.

Unsafe drinking water almost 15 minutes every day.

5,000 incorrect surgery operations per week.

2 short or long landings at most major airports each day.

200,000 incorrect drug prescriptions each year.

No electricity for almost 7 hours each month.

For many people, 99% good indicates a quality product or service; but this list might change that thinking.

Source: Motorola

Is Six Sigma software good enough?

Today, software comes in many forms and is an integrated part of many products. There is operating system software, application software, and the user developed software for a specific process. With all this software it is possible to have several million lines of code controlling a process, so it begs the question; is 6σ quality (3.4 defects per million lines of code) acceptable?

Software developers reading about concepts of process capability, often conclude it does not apply to them. Programmers often don’t believe they have specification limits.

Dr. Terry Heng, Motorola’s vp of corporate software development explained that good engineering practices should be applied to software, just like they are to other disciplines. By following standards, tools, and techniques, promoted by the Software Engineering Institute (SEI) of Carnegie Mellon University (Pittsburgh, Pa.), development of reliable software is possible. But Dr. Heng cautions, ‘Depending on how fast the software is needed will impact the quality of the software. Software produced on a fast track will be about 3σ. Software produced in about one-man year will be about 4σ; and software produced in about two-man years will be near 6σ.’

SEI ranks software development processes by levels. Software meeting SEI level 3 requirements is equal to about 5.8σ, level 4 software is about 6σ, and level 5 software is equivalent to 6.3σ. Software used on NASA’s space shuttle is at level 5. According to Dr. Heng, Motorola is producing software that meets SEI level 4, with some elements at level 5.

Getting the bugs out of software has always been a burden and usually is accomplished using regression testing. But with 10 million lines of code, regression testing can take a very long time. Dr. Heng shares, ‘Debugging tools are available that can produce level 3 (5.8σ) software, and it’s getting better. One of the ways Motorola manages software quality is through reusable modules. Once a module is designed, developed, tested, and certified, programmers are encouraged to use it ‘as is.’ ‘

Looking three or four years into the future, software development will continue to remove the human element through accelerated use of certified modules and automatic software layout and code development tools. ‘There will be those who argue that automatic tools do not write efficient software, and that’s true. But allowing humans to get in and ‘tweek’ the software for a 5% improvement is just not worth the risk,’ explains Dr. Heng.

So, is 6σ software good enough? Not if the software is used in medical diagnostic equipment, airplane flight control systems, industrial safety systems, or when billing my credit card.

For more information about the Software Engineering Institute, visit

CompactPCI systems approach Six Sigma capability

Using operating system software from Enea OSE Systems (Dallas, Tex.), Motorola Computer Group (Tempe, Ariz.) is delivering the CPX8000 family of CompactPCI systems with 99.999% (about 5.7σ) availability. This level of availability is made possible by the redundancy of every active component, enabling any active module to be exchanged for repair or upgrade while the system continues to operate. Redundancy extends to system-slot processor boards. Two CPU (central processing unit) boards can operate together to share workload, with fallback to a single CPU in the event of failure, or alternatively one CPU board can operate in standby mode, ready to take over if the active CPU fails. OSE’s real-time operating system (RTOS) is designed for stringent requirements of high-availability and fault-tolerant applications. Software extensions have been added to RTOS to ensure hardware and software hot-swap capability, and to assist in quickly deploying fault-tolerant systems.

OSE’s link handler allows a team of designers to implement interprocess communication regardless of whether the application uses a single or multiple CPUs. With link handler, communications between application processes is transparent, even when the processes are located on different system nodes.

OSE’s program handler supports loading, changing, or removal of programs during run-time. Built-in supervision features allow processes within an application to be alerted if a program has been removed or when it reappears. OSE’s message-passing architecture offers large applications, such as distributed telecommunications systems, fault-tolerance with extremely efficient signal handling that results in high data throughput.

For more information about OSE’s RTOS, visit