1.2 What Is Six Sigma?
If you are familiar with this term and methodology, you may want to skip this section. Readers new to this business strategy—and it is a business process—will want to read this section carefully, keeping in mind that it is only an introduction.
Six Sigma can be read about and still greatly misunderstood. In 2006, a shareholder of Honeywell proposed an item to be voted upon at the annual shareholder meeting that Honeywell drop a related use of the term as misleading stockholders and the general public. There are many detailed and descriptive texts on the subject so we will attempt to be brief here, mostly for the benefit of any readers who are new to these topics. I will attempt to describe Six Sigma in terms of its three most widely applied and published interpretations, namely as a metric, a project methodology, and finally as a company initiative.
1.2.1 Six Sigma: The Metric
We will describe Six Sigma as a metric in a very brief manner here, as other sources contain more-detailed descriptions.4,5 When Six Sigma was developed at Motorola in the 1980s, products were measured in terms of quality by counting their defects. For brevity's sake, we will define a defect for our purposes as anything not meeting customer expectations or requirements. Suppose for example that a customer requires on-time deliveries, specified as "no sooner than two days before promise date nor later than 1 day after promise date." Any delivery made to that customer outside of that range is unacceptable, considered to have a delivery defect. Other examples include packing defects, such as missing or incorrect items; shipping defects, such as arrival with damage or missing paperwork; and, of course, product defects in performance or appearance. All defects should be counted if they create customer dissatisfaction or add cost to the business for inspection and correction. However, not all defects are equal: Some have more impact on customer satisfaction and cost than do others. Sound business prioritization must take precedence over simple defect counting as if all were equal.
The goal of Six Sigma can be stated as simple as: "to define, measure, analyze, improve, and control the sources of variations that create defects in the eyes of customers and the business." The count of defects is a key metric. In fact, Motorola production operations focused for a time almost exclusively on reducing defects by an order of magnitude over a specified period of time. If we declare that any defective item has at least one defect, then we can (loosely) state that the yield of any ongoing process is equal to 1 minus the sum of the defects created in that process:
Yield = 1 - S(defects)
We can also count opportunities to create defects in a process. This might be as simple as counting the number of major steps in the process or as complex as looking in-depth at the number of opportunities for errors, mistakes, and non-conformities in the entire process. Whatever method is used, it is prudent to standardize and not change it. For assembled products, one of the best and simplest methods I have seen is to count the number of components for a product and multiply the total by three. The logic of this method is that each component can (1) be purchased correctly or incorrectly, (2) be assembled correctly or incorrectly, and then (3) perform correctly or incorrectly. Each purchased subassembly counts as only one component.
To normalize across different products and processes, the sigma metric involves counting defects and opportunities for defects, and then calculating a Sigma Value. An interim step is to calculate the Defects per Million Opportunities (DPMO) as:
DPMO = S(defects) ÷ S(opportunities) x 106
We use DPMO to arrive at the Sigma Value. The higher the Sigma Value, the higher the quality. The table in Figure 1-5 equates Yield, equivalent defects per million opportunities or DPMO, and the Sigma Value.
Figure 1-5. Yield, DPMO, and Sigma Values
99.99966% Yield 3.4 DPMO 6 Sigma
99.9767% Yield 233 DPMO 5 Sigma
99.379% Yield 6210 DPMO 4 Sigma
93.32% Yield 66807 DPMO 3 Sigma
69.1% Yield 308,537 DPMO 2 Sigma
We can see that when yields are in the range of 60 percent to 95 percent, we need few other measures to view improvement. However, as we approach 95 percent, our yield measure becomes fairly insensitive and it makes sense to find a more sensitive metric—hence Motorola's use of defect counting. If we further need to compare across vastly different product and process complexities, it makes sense to normalize by opportunity counts. Ultimately, as there are fewer defects to count, it is more important to move away from counting flaws to continuous-scaled measures of performance that tie well to customer satisfaction and product value.
So how do different products and services compare on Sigma Values? Figure 1-6 illustrates DPMO levels in parts-per-million (PPM) versus the Sigma Value.
Figure 1-6 Sigma Value versus DPMO, Scaled in PPM
How is the Sigma Value calculated? In short, it is the number of standard deviations from the mean of an equivalent normal distribution from a specified limit having the same defect rate as the one reported as DPMO. If we assume a defect occurs when any output is beyond the specification limit, then the odds of a defect at a Sigma Value of 6 are approximately 1 in 1 billion. However, defect measurements made in the short term are not entirely correlated with results over the long term. Over time, if left unchecked, process quality degrades with shifts and drifts in the average output value, inflation of the Sigma Value, or both. A rule of thumb is needed to account for output shifts or degradation over time. Though this is somewhat controversial, George Box has commented that he is glad some accounting for long-term shift is being done.
The rule that is usually applied is that over time, the average change is 1.5 standard deviations. This shift arises originally from work done on tolerance stack-ups and shifts.6,7,8 The switch to universal acceptance of this amount of shift is often based on empirical experience and not a theoretical construct.9 I have personally seen long-term shifts of the mean over time between 0 and 4 Sigma, and my best recommendation is to aim for 1.5 Sigma. If we reduce our Six Sigma distance by this amount, we get a 4.5 Sigma distance to the upper specification, and the resulting odds of being out of specification rise to 3.4 defects per million items. This is the defect rate number most often touted in Six Sigma quality descriptions. See Figure 1-7 for an illustration of the shift between short-term and long-term performance.
Figure 1-7 The Shift from Short-Term to Long-Term Performance
So now we have a Six Sigma metric description, albeit in an oversimplified way for those who are experienced in the craft. This metric can be used in many ways, from counting simple defects to performing simple measurements of product or service output. Knowing these measurements is a key step in delivering quality as measured in the eyes of customers.
1.2.2 Six Sigma: The Project Methodology
Six Sigma is more than just a metric; it is a project methodology as well. Often QFD is utilized as one of the tools within a Six Sigma project methodology. I view the difference between a tool and a methodology as follows:
- a methodology comprises several steps to achieve an aim or purpose, using multiple tools
- a tool comprises a single function, or multiple functions that may be applied in several ways
When Six Sigma is viewed as a process-improvement methodology, it typically follows a five-phase approach. This approach is sequenced as Define, Measure, Analyze, Improve, and Control (DMAIC). Every Six Sigma process-improvement project follows this sequence, although it may involve taking several loops back, as it is a discovery process in search of causes and controls.
Six Sigma earns its clout by completing business projects aimed at significant process problems. These problems are identified by an organization in two ways: either top down or bottom up. The bottom-up approach is to listen to customer complaints and/or list internal chronic problems. The top-down approach is to set aggressive business cost-reduction goals and then refine them to a manageable level. Projects are defined around these issues, and all follow the general phases DMAIC. For more on this "project-linked-to-business-goals" approach, see for example Six Sigma: The First 90 Days (2006) by Dr. Stephen Zinkgraf.10
Within each of the DMAIC phases, key tools are applied to solve the process problems encountered in the project. Project leaders are called either Black Belts or Green Belts in general. Black Belts usually lead the higher-value, higher-risk, higher-visibility, more-complex projects in an organization and receive extensive training in process-improvement tools and statistical analysis, along with some team-building and presentation training depending upon organizational needs. A Black Belt may be deployed in many different areas of the company to solve a process problem. Green Belts typically lead smaller projects, usually within their local areas of expertise, and receive less training. Industry body-of-knowledge standards exist for Black Belt practitioners; and these leaders typically receive four weeks of training spread out over four or five months. Green Belts typically receive nine or ten days of training in two or three sessions spread out over two or three months. Both types of training are excellent for developing your people into more valuable, more versatile company resources.
Much has been written on the tools topic, and an excellent practical text written by my colleague Dr. Ian Wedgwood is called Lean Sigma: A Practioner's Guide.11 This book is very straightforward regarding the tools involved in the methodology.
While consulting at Sylvania in 1998, I was challenged by some excellent engineers to explain how all the tools fit together. I started with a flipchart, and eventually put the major tools into a PowerPoint presentation that many have commented is helpful, so I will repeat it here as a series of figures.
The first step is to begin with the end in mind. We seek potential causes of our primary process issue (Y) and the relationship of the causes to Y. The issues may be multiple but we look at them separately and consider each issue or defect as a process result or Y. This can be stated mathematically as:
Y = f(x1, x2, x3, x4, . . ., xn)
Y can be a delivery time, a processing time, a product defect, a service error, an invoicing error or anything that we can explicitly state which can be measured. This Y = f(x) transfer function is the ultimate objective in all good product development work. It is necessary to move forward in QFD as we can associate the Y with Whats and the Xs with Hows. These transfer functions are needed to fill in each QFD matrix. Regarding the equation, some of the tools in the roadmap work on the Y, some help find Xs, and some help establish the relationships to the Y. In the Define and Measure phases we are applying tools at some point that work on Y. These tools are: Measurement System Analysis (MSA), Process Control Charts, and Capability Analysis. MSA helps us determine how much error exists in measuring Y. Control Charts tell us if the Y is stable or predictable, and Capability Analysis helps determine the extent of variation in Y relative to a requirement established during the Define phase. Together these two are called an Initial Capability Assessment. So now our picture looks like Figure 1-8.
Figure 1-8 Begin with the End in Mind,Y = f(x)
The next steps in the Measure and Analyze phases require identifying all the process Xs that are associated with this Y and then funneling them down to the potential few independent variables that truly influence Y. A Process Map is a tool to collect and identify these Xs, and a variant of a QFD matrix called the Cause and Effect matrix (C&E Matrix) combines the teams knowledge to "funnel" the larger list to known or observed influencers. A Failure Mode and Effects Analysis, or FMEA, helps establish how each independent X variable can be wrong, how severe is its impact, how often it occurs, and whether it can be detected. This helps make some "quick hits" to improving independent potential actors or Xs, and begins stabilizing the process. We now have a smaller number of variables, and we study these Xs passively (Multi-vari) to see if any have statistical correlation with actual variation observed in the Y variable, as shown in Figure 1-9.
Figure 1-9 Finding and Funneling Down the Independent Variables
In the Improve phase, we take the output of our funnel and begin Design Of Experiments (DOE) to establish true cause and effect relationships of Y = f(x) on our reduced list of suspect independent variables. From there we identify the true actors causing the variability in Y and establish proper controls with a control plan (Figure 1-10).
Figure 1-10 Establishing Y = f(x), Finding the Key Xs, and Setting Proper Controls
Now we have the major tools of the Process Improvement Methodology in sequence and relationships to establishing our goal of Y = f(x)!
Six Sigma applied to process problems has as its fundamental aim or goal to define the equation that describes where your desired output gets its dependency: i.e., Y = f(x). So when applying the tools our goal is to develop this key equation. In addition to the Six Sigma Tools, Lean Manufacturing tools were integrated as part of the combining of two separate initiatives while I was at AlliedSignal, now Honeywell, thanks to Larry Bossidy's wisdom on initiatives. There are certainly many ways to combine these initiatives. As a set of tools, it makes sense to combine them with the phased approach of DMAIC. Early on at SBTI, I wrote the DMAIC tools for Six Sigma in the form of the M-A-I-C loop shown in Figure 1-11. We then integrated the Lean tools into these four phases, as seen in Figure 1-12. Following this "Lean Sigma Roadmap" of tools within the DMAIC phases, one can work a project towards a result with less variation in process output, process times, and process flow. Please recall the earlier comments in the methodology introduction that each individual tool may be applied in many ways. This is just one approach to showing how Lean tools and Six Sigma tools may be integrated into a unified methodology of MAIC.
Figure 1-11 Six Sigma Tools in the M-A-I-C Sequence
Figure 1-12 Six Sigma and Lean Tools in the M-A-I-C Sequence
1.2.3 Six Sigma: Deploying as an Initiative
To begin a Six Sigma initiative, top management or executive leaders must create the vision of what the initiative is to do for the company, customers, employees, and owners. They need to foster acceptance with communication, involvement, and leading from the front of the deployment. They must set goals, timing, and financial expectations for the program if it is to be effective. These are often done in off-site meetings or executive workshops. The previously cited work by Zinkgraf12 provides more depth on these subjects. From these workshops, a deployment plan is developed for the Six Sigma initiative, and it is refined further by champions and sponsors of this effort. The refinement is around people, projects, and training timetables.
Refinement around people and projects means getting the right people on the bus, and then dropping them into the right project areas, armed with the right tools and training. To have great project work on significant business problems, a company needs sponsorship by the business ownership, management, or executive force, or all three depending upon the size of the organization. This requires initiative alignment with strategy, people, processes, business metrics, and customer needs. Alignment with customer needs, strategy, internal stakeholder needs, and business metrics is a great application of QFD and will be discussed further in Chapter 2.
The Black Belts or Green Belts need to know they have support and access to whatever is needed to conduct their project work. It is generally management's job to set expectations and then inspect for progress on tasks, measured against agreed-upon goals. Management must also support and coach the project leaders through the first few projects, and to do so must understand the basics of Six Sigma. People who fill this role in a Six Sigma initiative context are usually described and trained as either Champions or Sponsors. Champions or Sponsors identify the business areas and rough out the projects that are best linked to the business goals and needs. They also identify individuals for training as Green Belts and Black Belts. Green Belts and Black Belts take the projects to training and follow a Plan-Train-Apply-Review cycle through two, three, or four training sessions, depending upon the focus, level, and deployment plan within a company.
Due to the scope of a Six Sigma Initiative and the nature of this book, we will delve no further into the initiative topic. The previously cited work by Zinkgraf13 will provide more depth in this area.