Chapter 3: The Six Sigma DMAIC Model

Every methodology has a conceptual approach to work with. Six sigma is no different and, in fact, has two lines of approach. The first is to address existing problems, and the second, to prevent problems from happening to begin with. The six sigma methodology has adopted the old plan-do-study-(check)-act (PDS(C)A) approach, with some very subtle variations in that breakthrough strategy. This approach is a functional one—it clearly shows the correct path to follow once a project has been selected. In its entirety, the approach is the define, measure, analyze, improve and control (DMAIC) approach.

The stages of the DMAIC model

Define

The first stage—define—serves as the platform for the team to get organized, determine the roles and responsibilities of each member of the team, establish team goals and milestones and review the process steps. The key points to be defined at this stage are the voice of the customer, the scope of the project, the cause and effect prioritization (a list that the team creates for pursuing the specific project based on cause and effect criteria) and project planning. (aligning to the business strategy and the preliminary definition of the project).

Each of these points can be linked to the customer (some obviously and others not so), and it is essential to appreciate and understand this link to the customer before and during this stage of the model. The following are the steps to take to complete the define phase of the DMAIC model:

• Define the problem. The problem is based on available data, is measurable and excludes any assumptions about possible causes or solutions. It must be specific and attainable.

• Identify the customer. This is more demanding as we systematically begin the process of analysis. We must identify who is directly impacted by the problem and at what cost. We begin by conducting a random sample analysis to identify the overall impact and then we proceed with a detailed analysis of the cost of poor quality (COPQ). The focus of the team here is to identify a large base of people affected by poor quality.

• Identify critical to quality (CTQ) characteristics. By identifying CTQ characteristics, the project team determines what is important to each customer from the customer's point of view. Identification of CTQ characteristicts ascertains how these particular features appear when meeting customer expectations. Typical questions here are: What is "good condition?" and What is "on time?"

• Map the process. Mapping of the process in this stage of the define phase of the six sigma methodology is nothing more than a high level visual representation of the current process steps leading up to fulfillment of the identified CTQ characteristics. This "as is" process map will be useful throughout the process as:

  • A method for segmenting complex processes into manageable portions.

  • A way to identify process inputs and outputs.

  • A technique to identify areas of rework.

  • A way to identify bottlenecks, breakdowns and non-value-added steps.

  • A benchmark against which future improvements can be compared with the original process.

  Any organization is a collection of processes, and these processes are the natural business activities you perform that produce value, serve customers and generate income. Managing these processes is the key to the success of the organization. Process mapping is a simple yet powerful method of looking beyond functional activities and rediscovering core processes. Process maps enable you to peel away the complexity of your organizational structure and focus on the processes that are truly the heart of your business. Armed with a thorough understanding of the inputs, outputs and interrelationships of each process, you and your organization can understand how processes interact in a system, evaluate which activities add value for the customer and mobilize teams to streamline and improve processes in the "should be" and "could be" categories. It should be noted that understanding the process is an important objective of the process map. However, something that is just as important, and usually undervalued from constructing a process map, is the benefit of the alignment of the team to the process at hand. Once this alignment occurs, and everyone in the team understands what is expected, the conclusion of a successful project is a high probability.

• Scoping the project. The last step of the define stage is scoping the project and if necessary, updating the project charter. During this step the team members will further specify project issues, develop a refined problem statement and brainstorm suspected sources of variation. The focus of this step is to reduce the scope of the project to a level that ensures the problem is within the team's area of control, that data can be collected to show both the current and improved states and that improvements can be made within the project's timeframe.

At the end of this stage, it is not uncommon to revisit the original problem statement and refine it in such a way that the new problem statement is a highly defined description of the problem. Beginning with the general problem statement and applying what has been learned through further scoping, the team writes a refined problem statement that describes the problem in narrow terms and indicates the entry point where the team will begin its work. In addition, a considerable amount of time is taken at this step to identify the extent of the problem and how it is measured.

Ultimately, the purpose of this stage is to set the foundations for the work ahead in solving a problem. This means that an excellent understanding of the process must exist for all team members, as well as complete understanding of the CTQ characteristics. After CTQ factors are identified, everyone in the team must agree on developing an operational definition for each CTQ aspect. Effective operational definitions:

• Describe the critical to quality characteristics accurately.

• Are specific so that the customer expectation is captured correctly.

• Are always written to ensure consistent interpretation and measurement by multiple people.

Whereas typical methods of identifying CTQ characteristics include but are not limited to focus groups, surveys and interviews, the outputs are CTQ characteristics, operational definitions and parameters for measuring.

Measure

The second stage of the DMAIC model—measure—is when the team establishes the techniques for collecting data about current performance that highlights project opportunities and provides a structure for monitoring subsequent improvements. Upon completing this stage, we expect to have a plan for collecting data that specifies the data type and collection technique, a validated measurement system that ensures accuracy and consistency, a sufficient sample of data for analysis, a set of preliminary analysis results that provides project direction and baseline measurements of current performance.

The focus of this stage is to develop a sound data collection plan, identify key process input variables (KPIV), display variation using Pareto charts, histograms, run charts, and baseline measures of process capability and process sigma level. The steps to carry through this stage are:

• Identify measurement and variation. The measure subsets establish the requirements of measurement and variation, including: a) the types and sources of variation and the impact of variation on process performance, b) the different types of measures for variance and the criteria for establishing good process measures, and c) the different types of data that can be collected and the important characteristics of each data type. As part of this step the types of variation must be defined. There are two types of causes of variation:

  • Common causes. These are conditions in a process that generate variation through the interaction of the 5Ms (machine, material, method, measurement, manpower) and 1E (environment). Common causes affect everyone working in the process, and affect all of the outcomes. They are always present and thus are generally predictable. They are generally accepted sources of variation and offer opportunities for process improvement.

  • Special causes. These are items in a process that generate variation due to extraordinary circumstances related to one of the 5Ms or 1E. Special causes are not always present, do not affect everyone working in the process and do not affect all of the outcomes. Special causes are not predictable.

• Determine data type. In this step the team must be able to answer the question, "What do we want to know?" Reviewing materials developed during the previous stage, the team determines what process or product characteristics they need to learn more about. A good start is the definition of the data type. This is determined by what is measured. Two types of data can be collected by measuring:

  • Attribute data. One way to collect data is to merely count the frequency of occurrence for a given process characteristic (e.g. the number of times something happens or fails to happen). Data collected in this manner is known as attribute data. Attribute data cannot be meaningfully subdivided into more precise increments and is discrete by nature. "Go/no go" and "pass/fail" data are examples of this category.

  • Variable data. A different way to look at data is to describe the process characteristic in terms of its weight, voltage or size. Data collected in this manner is known as variable data. With this type of data, the measurement scale is continuous-it can be meaningfully divided into finer and finer increments of precision.

• Develop a data collection plan. In developing and documenting a data collection plan the team should consider:

  • What the team wants to know about the process.

  • The potential sources of variation in the process (Xs).

  • Whether there are cycles in the process and how long data must be collected to obtain a true picture of the process.

  • Who will collect the data.

  • How the measurement system will be tested.

  • Whether operational definitions contain enough detail.

  • How data will be displayed once collected.

  • Whether data is currently available, and what data collection tools will be used if current data does not provide enough information.

  • Where errors in data collection might occur and how errors can be avoided or corrected.

• Perform measurement system analysis. This step involves performing graphical analysis and conducting baseline analysis. During this step, the team verifies the data collection plan once it is complete and before the actual data is collected. This type of analysis is called a measurement system analysis (MSA). A typical MSA indicates whether the variation measured is from the process or the measurement tool. The MSA should begin with the data collection plan and should end when a high level of confidence is reached that the data collected will accurately depict the variation in the process. By way of a definition, MSA is a quantitative evaluation of the tools and processes used in making data observations. Perhaps the most important concept in any MSA study is that if the measurement system fails to pass analysis before collecting data, then further data should not be collected. Rather, the gauge should be fixed, the measurement system should be fixed and the measurement takers should be trained.

• Collect the data. During this step, the team must make sure that the collected data is appropriate, applicable and accurate, and that it provides enough information to identify the potential root cause of the problem. It is not enough to plan carefully before actually collecting the data and then assume that everything will go smoothly. It is important to make sure that the data continues to be consistent and stable as it is collected. The critical rules of data collection are:

  • Be there as the data is collected.

  • Do not turn over data collection to others.

  • Plan for data collection, design data collection sheets and train data collectors.

  • Stay involved throughout the data collection process.

  The outcome of this step must be an adequate data set to carry into the analyze stage.

Analyze

The third stage—analyze—serves as an outcome of the measure stage. The team at this stage should begin streamlining its focus on a distinct group of project issues and opportunities. In other words, this stage allows the team to further target improvement opportunities by taking a closer look at the data. We must remember that the measure, analyze and improve stages quite frequently work hand in hand to target a particular improvement opportunity. For example, the analyze stage might simply serve to confirm opportunities identified by graphical analysis in the measurement stage. Conversely, the analyze stage might uncover a gap in the data collection plan that requires the team to collect additional information. Therefore, the team makes sure the appropriate recognition of data is given and applicable utilization is functional, as well as correct. Yet another important aspect of this stage is the introduction of the hypothesis testing for attribute data. On the other hand, in the case of variable data we may want to use: analysis of means (1 sample t-test or 2 sample t-test), analysis of variance for means, analysis of variance (F-test, homogeneity of variance), correlation, regression and so on.

At the end of this stage the team should be able to answer the following questions:

• What was the improvement opportunity?

• What was the approach to analyzing the data?

• What are the root causes contributing to the improvement opportunity?

• How was the data analyzed to identify sources of variation?

• Did analysis result in any changes to the problem statement or scope?

We are able to do this by performing the following specific sequence of tasks:

• Perform capability analysis. This is a process for establishing the current performance level of the process being. This baseline capability will be used to verify process improvements through the improve and control phases. Capability is stated as a short-term sigma value so that comparisons between processes can be made.

• Select analysis tools. This step allows the team to look at the complete set of graphical analysis tools to determine how each tool may be used to reveal details about process performance and variation.

• Apply graphical analysis tools. This refers to the technique of applying a set of basic graphical analysis tools to data to produce a visual indication of performance,

• Identify sources of variation. This refers to the process of identifying the sources of variation in the process under study, using statistical techniques, so that significant variation is identified and eliminated.

The analyze stage continues the process of streamlining and focusing that began with project selection. The team will use the results produced by graphical analysis to target specific sources of variation.

As an outcome of the analyze stage, the team should have a strong understanding of the factors impacting their project including:

• Key process input variables (the vital few Xs that impact the Y).

• Sources of variation—where the greatest degree of variation exists.

Improve

The fourth stage—improve—aims to generate ideas; design, pilot and implement improvements; and validate the improvements. Perhaps the most important items in this stage are the process of brainstorming, the development of the "should be" process map, the review and/or generation of the current FMEA (failure mode and effect analysis), a preliminary cost/benefit analysis, a pilot of the recommended action and the preliminary implementation process. Design of experiments (DOE) is an effective methodology that may be used in both the analyze and improve stages. However, DOE can be a difficult tool to use outside a manufacturing environment, where small adjustments can be made to input factors and output can be monitored in real time. In non-manufacturing, other creative methods are frequently required to discover and validate improvements.

The following steps should be taken at this stage:

• Generate improvement alternatives. The emphasis here is to generate alternatives to be tested as product or process improvements. The basic tools to be used here are brainstorming and DOE. With either tool, a three-step process is followed:

  1. Define improvement criteria—develop CTQ characteristics.

  2. Generate possible improvements—the best potential improvements are best evaluated based on the criteria matrix.

  3. Evaluate improvements and make the best choice.

As a result of these steps, several alternatives may be found and posted in a matrix formation. The matrix should have at least the following criteria: "must" criteria (the basic items without which satisfaction will not occur) and "desirable" criteria (items that are beyond the basic criteria and do contribute to performance improvement). Once these are identified a weight for each is determined, either through historical or empirical knowledge, and appropriately posted in the matrix. At that point each criteria is cross-multiplied by the weight and the appropriate prioritization takes place. This is just one of many prioritization methods. Other prioritization methods may be based on cost, frequency, effect on customer and other factors.

• Create a "should be" process map. This map represents the best possible improvement the project team is able to implement. It is possible that a number of changes could be made to improve a process. The individual process map steps will serve as the input function of the FMEA.

• Conduct FMEA (failure mode and effect analysis). The FMEA is meant to be a "before the failure" action, not an "after the fact" reaction. Perhaps the most important factor in any FMEA is the fact that it is a living document and therefore it should be continually updated as changes occur or more information is gained.

• Perform a cost/benefit analysis. This analysis is a structured process for determining the trade-off between implementation costs and anticipated benefits of potential improvements.

• Conduct a pilot implementation. This step is a trial implementation of a proposed improvement, conducted on a small scale under close observation.

• Validate improvement. One of the ways to validate the effectiveness of the changes made is to compare the sigma values before and after the changes have been made. Remember, this means to compare the same defects per million opportunity.

Control

The fifth stage—control—is to institutionalize process or product improvements and monitor ongoing performance. This stage is the place where the transition from improvement to controlling the process and ensuring that the new improvement takes place. Of course, the transition is the transferring of the process from the project team to the original owner. The success of this transfer depends upon an effective and very detailed control plan. The objective of the control plan is to document all pertinent information regarding the following:

• Who is responsible for monitoring and controlling the process.

• What is being measured.

• Performance parameters.

• Corrective measures.

To make the control effective, several factors must be identified and addressed. Some of the most critical are:

• Mistake-proofing. This is to remove the opportunity for error before it happens. Mistake-proofing is a way to detect and correct an error where it occurs and avoid passing the error to the next worker or the next operation. This keeps the error from becoming a defect in the process and potentially impacting the customer CTQ characteristics.

• Long-term MSA (measurement system analysis) plan. Similar to the original MSA conducted in the measure stage, the long-term MSA looks at all aspects of data collection relating to the ongoing measurement of the Xs and high level monitoring of the Ys. Specifically, the long term MSA documents how process measurements will be managed over time to maintain desired levels of performance.

• Appropriate and applicable charts (statistical process control). A control is simply a run chart with upper and lower control limit lines drawn on either side of the process average. Another way to view the control chart is to see it as a graphical representation of the behavior of a process over time.

• Reaction plan. A reaction plan provides details on actions to be taken should control charts indicate the revised process is no longer in control. Therefore, having a reaction plan helps ensure that control issues are addressed quickly and that corrective actions are taken.

• The new or revised standard operating procedures (SOPs). Updating SOPs and training plans is the practice of revising existing documentation to reflect the process improvements.

At the end of the control stage, the process owner will understand performance expectations, how to measure and monitor Xs to ensure performance of the Y, and what corrective actions should be executed if measurements drop below the desired and anticipated levels. Furthermore, the team is disbanded while the black belt begins the next project with a new team.

Typical tools and deliverables for each of the stages of the DMAIC model are shown in Table 3.1

Table 3.1: Typical tools/methodologies and deliverables for the DMAIC model

Stage	Tools/methodologies	Deliverables
Define	Brainstorming Cause and effect diagram Process mapping Cause and effect matrix Current failure mode and effect analysis (FMEA) Y/X diagram CT matrix	The real customers Data to verify customers' needs collected Team charter—with emphasis on: problem statement project scope projected financial benefits High-level process map—"as is"
Measure	Process mapping Cause and effect FMEA Gauge R&R (repeatability and reproducibility) Graphical techniques	Key measurements identified Rolled throughput yielded Defects identified Data collection plan completed Measurement capability study completed Baseline measures of process capability Defect reduction goals established
Analyze	Process mapping Graphical techniques Multi-vari studies Hypothesis testing Correlation Regression	Detailed "as is" process map completed The sources of variation and their prioritization SOPs reviewed Identify the vital few factors KPIVs with appropriate and applicable data to support such KPIVs (Key process input variables) Refined problem statement to the point where the new understanding is evident Estimates of the quantifiable opportunity represented by the problem
Improve	Process mapping Design of experiments Simulation Optimization	Alternative improvements Implementation of best alternative for improving the process "Should be" process map developed Validation of the improvement—especially for key behaviors required by new process Cost/benefit analysis for the proposed solutions Implementation plan developed—a preliminary preparation for the transition to the control stage Communication plan established for any changes
Control	Control plans Statistical process control Gage control plan Mistake-proofing Preventive maintenance	Control plan completed Evidence that the process is in control Documentation of the project Translation opportunities identified Systems and structures changes to institutionalize the improvement Audit plan completed