Software Engineering Using Six Sigma
Matthew Stangel
Department of Software Engineering
University of Wisconsin – Platteville
Abstract
Six Sigma is a process for developing and improving other engineering processes. It was developed in the late 1980’s by Motorola to measure defects in an attempt to decrease their frequency. Since its creation, it has been adopted by multiple other large companies. My paper will discuss the basics behind what Six Sigma is, how it is used, why companies use it, and list some of the companies that have used Six Sigma. The paper will describe the two basic Six Sigma process variations and what situations they are used in, as well as how they can be used in Software Engineering. Finally, the paper will describe some criticisms of the Six Sigma process and reasons it might not be useable by everyone.
Introduction
Definition of Six Sigma
In the most basic of definitions, Six Sigma is a process for improving other engineering processes. The way Six Sigma accomplishes this improvement is by reducing the number of defects that exist in a product by attempting to improve the process that creates that product.Specifically, it aims to enforce at least 6 standard deviations between the mean and either specification limit for the results of a process as shown on a normal distribution curve. This equates to approximately 3.4 defects per million opportunities.These 6 standard deviations (or “sigmas”) are what give Six Sigma its name and are the basis behind the quality Six Sigma attempts to achieve.
To better understand some of the concepts of Six Sigma, definitions of key phrases are quite important. As stated in the previous paragraph, Six Sigma can be described as having 3.4 defects per million opportunities. In Six Sigma, a “defect” is defined as any place an undesirable result is produced or a customer specification is not met. Six Sigma measures how many defects can happen based on the number of “opportunities” a defect can occur at. In Six Sigma, a defect“opportunity” is any place within a product that a defect can occur. This includes places such as requirements and interface design as well as bugs in the product.
History of Six Sigma
Six Sigma was developed in the late 1980’s at Motorola. Motorola had been experiencing troubles with their product quality in the late 1970’s[5], and they decided that they needed to improve their product development to raise their quality. In 1986, one of Motorola’s employees by the name of Bill Smith finished developing a process to reduce defects in products. He proposed that most defects that existed in products were either “early life failures”, or problems latent in the design of the product, or “wear out failures”, or problems that developed once a product had been used extensively and its components worn down. He symbolized these most often occurring offailures with the “bathtub curve”, as the graph of when defects occurred within the products he examined formed a curved-sided shape similar to a bathtub (as shown in Figure 1). The wear out failures were considered mostly unavoidable, but the early life failures were simply problems with design quality and thought to have been avoidable with more careful product design. This concept eventually led to his creation of Six Sigma to address these problems.
Figure 1: The Bathtub Curve[5]
Six Sigma Theory
As stated earlier, Six Sigma holds its quality standard at having six standard deviations between the mean and nearest specification limit of a process output. However, it may be hard to visualize exactly what this means and why six is the number of sigmas the process holds to be standard. To start, the six sigmas between the mean and nearest specification limit cause almost none of the normal curve that symbolizes the production output to be outside of the specification limit. In fact, so little of the graph is outside of the specification limits that it may cause some concern that the standard of six sigmas is actually too harsh and creating products of that quality level is almost unnecessary except in special circumstances and highly critical products. However, this extra caution in specification boundaries allows for change in the product over time. When he created Six Sigma, Bill Smith realized that setting the minimum acceptable level of quality as a standard would not be able to adapt to any problems that the product may encounter over time. To combat this, he added a 1.5-sigma shift into Six Sigma. In other words, the mean of the production graph can move up to 1.5 sigmas to either side and still be within acceptable quality levels (as shown in Figure 2).
Figure 2: A normal curve showing Six Sigmas and including the 1.5 Sigma shift[5]
“Acceptable quality levels” may sound like a vague term, and speaking of production in terms of standard deviations may seem a little abstract. However, it is possible to calculate more concrete statistics using these figures. By computing the percentage of area under the curve within the specification limits, we can calculate some meaningful numbers based off of measurements of real life occurrences, as shown below.
Table 1: Real World Performance Levels
Using Six Sigma
Six Sigma Methodologies
In order to achieve its quality goals, Six Sigma employs two main methods of process improvement. Those two methods are called DMAIC and DMADV. While these two methodologies both have the same basic goal of making a process Six Sigma quality and also have some similarities in steps of execution, they are different in how they are used. DMAIC, which stands for Define, Measure, Analyze, Improve, and Control, is aimed specifically at improving existing processes by applying Six Sigma to them. Conversely, DMADV, which stands for Define, Measure, Analyze, Design, Verify, is structured to create entirely new processes with Six Sigma already integrated in them.
DMAIC
DMAIC is the Six Sigma method for improving existing processes to increase their quality to Six Sigma standards. Presumably, a process that DMAIC is applied to is not yet at Six Sigma levels of quality; otherwise, DMAIC will have no effect on the quality level of the output of that process.DMAIC has five phases, which are applied to a process sequentially.
Define
The first step of DMAIC is define. Define is the phase in which the project is defined. First, the project is identified based on the desired objectives of the process, including defining who the customers are and what they expect of the product produced by this process. This helps to define important aspects of the project such as what is considered a defect.Then, the aspects of the process that are critical to the quality of the product are identified to better locate the places in the process that are most important and should receive the most attention.
Measure
The second step of DMAIC is measure. In the measure step of DMAIC, the basic goal is to determine how to measure the problems in the process. Specifically, the major goal of this phase is to determine how to measure defects in the product, as it is difficult to fix a process if the problems with that process are not defined. In addition, in the measure phase, the inputs and outputs of the process as well as other aspects of the process that are critical to the product’s quality must be identified. Once these things are defined, acceptable parameters for each of these parts of the process can be decided as well as the potential problems created if the product does not meet those parameters. This is also the phase in which problems occurring in existing product are measured to better understand what problems currently exist in the product to better understand how they should be avoided during future production.
Analyze
In the analysis phase, the aspects of the process measured during the measure phase are analyzed so that they may be used to improve the process. First, the difference between the current product specifications and the desired specifications must be found. This helps identify the most likely causes of the problems in the current process so they can be fixed. Then, how to fix the problems identified is determined. In addition, the potential sources of problems in the process should be prioritized in order to be able to fix the most likely causes of problems first.
Improve
The improve phase is the phase in which changes identified in the first three phases are implemented so that the improvements identified can take effect. Solutions to the various problems previously identified are created and then applied to the process. An additional part of this phase is determining the effects of not implementing the improvements as well as taking too long to implement the improvements, as sometimes not all improvements can be applied quickly or occasionally cannot be applied at all due to resource restrictions
Control
The last phase of DMAIC, control, is about controlling the process so that the improvements made on the process continue to affect the process and have the desired effect. Since by this phase the changes are already designed and implemented, all that really remains is to monitor the effects of the changes. Most importantly, this phase involves verification that the changes made to the process have the desired and expected result.
DMADV
DMADV, while in some ways similar to DMAIC, has a different purpose. While DMAIC is designed to improve an existing process to give it Six Sigma levels of quality, DMADV is used when creating a new process to ensure that it has Six Sigma quality from the beginning. It can also be used when existing DMAIC-modified processes still do not meet the Six Sigma level of quality and a new process needs to be made to replace it. The fact that it is used when creating new processes gives it the additional name Design for Six Sigma. Like DMAIC, DMADV also has five steps, and three of them are basically identical to the ones in DMAIC.
Define
The first phase of DMADV is the same as DMAIC – define. The define phase of DMADV is identical to the define phase of DMAIC. Everything that happens in the DMAIC define phase happens the same way in the DMADV define phase, except that the process is still being designed.
Measure
The second phase of DMADV is also the same as DMAIC – measure. However, the fact that a process does not already exist to be measured changes a few key points of this phase. Rather than measure the key points of the process and measure the defects, the customer specifications are measured to better design a process to meet those specifications.
Analyze
The third phase of DMADV is the last phase it shares with DMAIC – analyze. Like measure, the DMADV analyze is similar to the DMAIC analyze, but a few key differences exist because a new process is being created instead of an old process being repaired. Rather than analyzing the defects in the current product and process, the options for meeting customer needs are analyzed.
Design
The fourth phase of DMADV – design – is where it starts to be completely different from DMAIC. The three phases prior to design should have gathered enough information by this point that the process can begin to be actually designed, which is the main goal of this phase. The process is designed during this phase based on the previous phases so that it created products that meet customer specifications.
Verify
The final phase of DMADV is verify. In the verify phase, it is verified that the design created in the previous step produces the correct output and meets customer specifications. This determines whether or not the process was successful in creating a Six Sigma process.
Using Six Sigma with Software
Six Sigma does not apply to writing code – it applies to software design instead. In fact, coding applies only to the improve phase of DMAIC and the design phase of DMADV. Six Sigma methodologies can be applied to software because software design has the ability to be measured. A study identified “around forty (40) metrics that can potentially be measured, and thus provide the basis for the execution of a defined and statistically based improvement methodology”[2]. Because software processes can be measured, they can have the steps of Six Sigma applied to them.
Learning Six Sigma
To learn how to apply Six Sigma techniques to processes, a person must undergo Six Sigma certified training. This is a multi-stage process, and as a person increases in Six Sigma proficiency, they receive different titles relative to their proficiency level. These titles are names “belts”, similar to the system or belts used to gauge martial arts proficiency.
Yellow Belts
Anyone involved in a Six Sigma process that does not have official training is referred to as a yellow belt. As yellow belts have no official training in Six Sigma, their roles in applying Six Sigma tend to be basic. A person only requires to have the knowledge that they are working on a Six Sigma project to be a Yellow Belt.
Green Belts
The first level of certification gives a person the title of Green Belt. Green Belts have basic knowledge of the inner workings of how to apply Six Sigma and tend to be in charge of any Yellow Belts working on the project. As it is least involved, it is also the least costly level of actual Six Sigma training.
Black Belts
Black Belts receive more training than Green Belts. As such, they are usually in charge of groups of Green Belts working on the project. Black Belts receive extensive training about the use of Six Sigma and tend to lead small projects. Black Belt certification is more costly than Green Belt, and in addition Green Belt certification must be taken before a person can be certified as a Black Belt.
Master Black Belts
The top stage of Six Sigma certification is a Master Black Belt. Master Black Belts not only have knowledge of all things Six Sigma, but also have the ability to train others in the use of Six Sigma. In large projects, Master Black Belts lead all of the Black Belts working on the project. Master Black Belt certification requires the largest monetary investment of the Six Sigma levels.
Champions
Champions are not really a level of Six Sigma certification, but their role is not to apply Six Sigma to a process. Instead, the person who sponsors the Six Sigma process is labeled as the Champion. The Champion is not even required to have any real involvement in the project other than sponsoring it, and that is generally all they do.
Results of Sigma
Benefits of Six Sigma
If Six Sigma is applied correctly to a process delivering a low quality product, the product should theoretically improve in quality once Six Sigma is applied to it. According to a report made by General Electric in 1996, “It has been estimated that less than Six Sigma quality, i.e., the three-to-four Sigma levels that are average for most U.S. companies, can cost a company as much as 10-15% of its revenues”[2].These figures result in an estimated $11.2-16.7 billion worth of savings for GE in 2001 based solely on the fact that they use Six Sigma[2]. Many other companies have seen similar results, as the list of successful companies that have used Six Sigma include Honeywell, Lockheed Martin, and Xerox.
Criticisms of Six Sigma
Six Sigma is not universally liked, as there are some criticisms against it. Perhaps the largest criticism against Six Sigma is that it is unoriginal. Critics of Six Sigma claim that it is simply a basic quality process with new terms attached to it. In addition, many critics point to the six standard deviations as being a completely arbitrary value to base quality on. They claim that it not only does not have any real meaning, but is not necessarily an applicable value to all situations. Another problem that Six Sigma presents is its cost. The high costs of training even one person to be Six Sigma certified is a very large investment for small companies.
Conclusion
Six Sigma is a tool to help low quality processes improve or to ensure processes are initially created with high quality. Six Sigma can be applied to aspects of Software Engineering, just like they can be applied to other kinds of engineering. It has proven to be an effective means to improve process quality, but it is costly and may not be for everyone.
References
[1] Adams, Cary W.; Gupta, Praveen (2003).“Six Sigma Deployment”. Burlington, MA: Butterworth-Heinemann
[2] Ideal Meadia LLC. i Six Sigma. Retrieved 1/17/2010. <
[3] Peterka, Peter. 2007. “The DMAIC Method in Six Sigma”. Retrieved 1/17/2010.
[4] Huesing, Tina. “Six Sigma Through The Years”. Motorola, 2008. Retrieved 1/17/2010.
[5] Peterka, Peter; Weed, Dr. Harrison; Wexler, Joseph; Jackson, Jeffery. Six Sigma us. Retrieved 1/17/2010. <