Six Sigma

Probably most popular approach to process improvement is Six Sigma. Since this short course is very limited in scope, we cannot address all of Six Sigma tools and techniques, but will highlight basic concepts behind Six Sigma. As we discussed earlier, we want to reduce variation (improve quality) to continuously improve our processes. Six Sigma provides a methodology for getting toroot cause of variation and reducing it. Despite what some might say, Six Sigma is not about forcing you to obtain perfection at any costs. It’s more about giving you a wide range of tools, applied in a discipline way for improving a process on a project by project basis. When applied toright kinds of projects, Six Sigma can yield significant results.

Project Selection

Six Sigma is executed through projects and since Six Sigma is very precise, it’s often better to start with smaller projects that have limited scope as opposed to large, organizational wide projects that are too difficult to manage. Additionally, projects need to have some justification behind being selected. So inworld of Six Sigma, it is very common to see a series of toll-gates or a formal business case to justifyproject. For example, projects will consume resources and time. There needs to be a clear payoff or return for doingproject. Additionally, it is useful to clearly defineexpected impact of projects and match these impacts against critical issues confrontingorganization. For example, a high level of customer complaints or product returns is a critical issue that might be ripe for a Six Sigma type project.

Five Phases - DMAIC

life cycle of six sigma work consists of five phases:

1. Define Opportunities: What must we do to meet VOC - Voice ofCustomer. In this phase, you must clearly identify your customers and analyze customer related information, translating this into Critical to Quality (CTQ). CTQ’s are requirements that your processes must perform up to if you expect to meet customer expectations. Once you understand this, then you can initiate six sigma projects to addressspecific performance issues.

2. Measure Performance: How much variation is taking place in our processes? In this phase, you will measure your variation in relation to an acceptable level of performance or specification limit. This is driven bycharacteristics of your CTQ. Certain statistical tools are used, such as sampling, frequency distribution, and control charts.

3. Analyze Opportunities: What areroot causes behind this variation? In this phase, you identifysources of variation. A good place to start is with a nonstatistical tool: Root Cause Analysis, includingFive Whys. Then you can begin to use certain statistical tools, such as Analysis of Variance, to better understandsources of process variation.

4. Improve Performance: What can we do to reduce this variation? vital few or root sources of variation are now identified. One ofmore popular tools used for improvement is called Design of Experiments (DOE).

5. Control Performance: How can we designprocess so that we never crossUpper or Lower Control Limits? This is where you sustain your desired performance levels and where practical, seek to improve it by removing more variation fromprocess.

The Basics

Sigma is a statistical measure of process capability in relation to how much deviation takes place inpopulation of data. It measuresvariability of the data. For every opportunity, there is a chance we might have a defect. This is typically expressed as Defects per Million Opportunities or DPMO. Defects representfailure to meet customer requirements. higher the sigma, more process outputs are able to meet customer requirements given fewer defects. following DPMO scale is used to expressdifferent sigma levels:



For various processes, we set targets which we will call "critical to" such as Critical to Quality (CTQ). This might be making pizzas in our pizza restaurant that are produced in 8 minutes. Each time we bake a pizza, there is some variation from this target of 8 minutes. If we plot each of these bake times, we can showdistribution on a graph. Additionally, our customers are willing to accept pizzas baked in 10 minutes, but likewise it takes us at least 6 minutes to put allingredients together for bakingpizza. These limits represent the Upper Specification Limit (USL) and Lower Specification Limit (LSL) within our distribution. goal is to "control" what happens within this range and when we bakepizza at exactly 8 minutes, we have Six Sigma quality - zero deviation from standard. As we get better and better at our baking process, we start to narrowrange, USL and LSL, so thatnormal distribution curve becomes tighter. This is how we express continuous improvement in the world of Six Sigma.

CTQ and VOC

Critical to Quality is customer driven and so we have to tap intocustomer to understand our requirements (CTQ). Six Sigma (as well as lean) requires that you are listening to the Voice ofCustomer or VOC. Inworld of Six Sigma, you are "insync" with VOC when you:

1. Provide a 100% solution tocustomer’s problem.
2. Minimal effort involved - not wastingcustomer’s time and efforts.
3. Giving the customer exactly what they need - no compromises.
4. Provide the value wherecustomer wants it.
5. Provide the value whencustomer wants it.
6. Compressdecision making process for the customer - make it easy for the customer to reachdecision.

This is perhaps one ofbiggest reasons why Six Sigma and Lean have become so popular - bar has been raised in terms of customer satisfaction. Additionally, any variation from the target increases costs. So Six Sigma is not just about improving quality and lowering costs, but also about customer satisfaction. Finally, there are two dimensions to CTQ - Customer driven CTQ’s coming from our external customers and process driven CTQ’s coming from our internal customers.

"There is a parable of the three blind men andelephant. Each is asked to identify what they are touching. first touches the tusk of elephant and identifies he is touching a spear. second touchestorso and claims what he is touching is a wall. third touchestail and think it’s a snake. This parable parallels Six Sigma. As its popularity has grown, different experts have marketed Six Sigma to fit their needs, not necessarily that of their customers. Of course, Six Sigma includes significant amounts of statistical tools. But many see Six Sigma as only statistics. They are wrong. Touch part ofwork that constitutes Six Sigma and it will look eerily similar to other quality approaches. Touch another part of Six Sigma and it only vaguely resembles a quality approach at all." - Six Sigma Execution by George Eckes

The Six Sigma Equation

Six Sigma begins with a simple equation that says - All outcomes areresult of inputs and process that acts on these inputs may introduce errors. Errors create variation and in the world of Six Sigma, variation is everything. This equation is expressed as:

Y = f (X) + E

Y: Desired outcome
f: Activities and Functions that convert inputs to outcomes
X: Inputs that are needed to producedesired outcome
E: Errors

If we go back to our pizza example, we bake pizzas with different outcomes or Y’s. Several different inputs are required before we can bakepizza - preparing the pie crust (input material), having cooks put all of the ingredients together (input labor) and using an oven (input equipment) to bakepizza. All of these inputs are X’s in our equation and we must measure these inputs (X’s) to get a profile of how our process performs in relation to our targeted performance.

Statistical Concepts

One ofattractions behind Six Sigma has to do with statistics. Statistics removes much of subjectivity that often plagues other forms of analysis. Opinions and speculation are replaced by applying statistical concepts to data. Some of these statistical concepts include:

1. Mean and Standard Deviation: Expressing process performance begins with the Mean and Standard Deviation. Mean representsaverage of your sample values; sum of all values divided by the number of observations in your sample. Standard Deviation isspread of data aroundmean. Standard Deviation is calculated by going through the following steps:

a. Calculatedifference from the mean for each observation.
b. Take square of each difference.
c. Sum all of your square values and divide by the number of observations less 1. NOTE: When calculatingstandard deviation for a sample (as opposed to the entire population), the number of observations is reduced by 1. This tends to improvecalculation so that the standard deviation of the sample is as close to the entire population as possible. It is rare that we will be measuringentire population.
d. Takesquare root of your value from step c (variance). This gives you the standard deviation.

Let’s go back to our pizza example. Suppose we made 6 observations of how long it takes to bake pizza. Our upper control limit is 8 minutes; i.e. we don’t want to take more than 8 minutes to bake pizzas. results of our six observations are:



2. Sigma Value: After calculatingMean and Standard Deviation, you need to express this performance related to CTQ - customer requirements. This is done by calculating the Sigma Value (sometimes called Z-Score) which represents the number of standard deviations frommean. However, in order for this to work we need a normal distribution of data. So it’s useful to do a histogram and plot your data, observing the curve or frequency distribution of your observations.

3. t test: Since we use samples to represent populations, we will most likely not know the standard deviation ofpopulation. And when our sample size is small (less than 30 observations), we can use t test to help us with a hypothesis test about the characteristics associated withpopulation.

4. F test: We may want to take samples from different segments ofpopulation, such as sampling only cheese pizzas, then sampling deluxe pizzas to see if this yields different results. You can use the F test to help understand differences in standard deviations between samples taken from different populations.

5. ANOVA: Used to conduct hypothesis testing when you have two or more groups of data. Like the t-test,purpose of ANOVA is to test the equality of the means betweendata groups. When you test and analyze only one variable (such as oven temperature in baking our pizzas), this is a One-Way ANOVA. If we tested two factors (such as oven temperature and dough texture of pizzas), this would be Two-Way ANOVA. testing of a combination of factors simultaneously in one test is referred to as a factorial experiment.

Design of Experiments (DOE)

number of inputs can be numerous (people, materials, equipment, technology, practices, methods, applications, etc.), making our six sigma equation look like:

Y = f (X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14, X15, X16, X17, X18, X19, X20)
What we really need to do is find out which of these inputs (X) is havingmost influence on our outcome (Y). By focusing on the "vital few" input variables, we gain control over the process. A series of different controlled experiments will get us tovital few. Experiments are managed based on:

1. Factors: possible X’s in our equation
2. Levels: range of values for each factor
3. Main Effects: change in Y from our experiment as we change our factor (X) from the lowest level tohighest level.
Factors areindependent variable and we want to quantify the impact on Y (response variable). In our pizza example, we might include these factors to help us understand variation in baking pizzas:



Each combination is an equation, contained within a matrix for all factors in our experiment. In order to getmost information, a full matrix is needed which contains all possible combinations of factors and levels. If this creates too many experimental runs, fractions ofmatrix can be taken.

"Probably few people know exactly what is meant by quality. Quality actually has different dimensions, which are all considered by consumers purchasing products. Although we as consumers may not know precisely what we mean by quality, we all recognize quality when we see it." - Myths of Japanese Quality by Ray and Cindelyn Eberts

Design for Six Sigma

"DMAIC" approach to Six Sigma seeks to improve existing processes. However, this is only half of six sigma management process. other half is to design and develop new processes to improve how we meet customer expectations. This is called Design for Six Sigma (DFSS). DFSS is used under two circumstances: Existing processes cannot be improved or a process to meet CTQ does not exist. Some oftools used in DFSS type projects include:

1. Quality Function Deployment (QFD): A methodology for identifying and categorizing customer requirements into a matrix. matrix prioritizes customer expectations on a scale from 1 (least important) to 5 (most important). Causeeffect requirements are also ranked; i.e. what is the correlation between a customer requirement and customer satisfaction. This is "roof" matrix that sits on top ofmain house matrix. Depending upon your approach, QFD may include several matrixes for capturing important relationships:



2. Failure Mode Effects Analysis (FMEA): Analytical approach directed toward problem prevention through which every possible failure mode is identified and risk rated. basic steps for FMEA are:

a. Identify various failure modes (spoiled materials, labor input mistakes, flawed method, equipment failure, etc.)
b. Identifyeffects
c. Determine the impact
d. Identify causes
e. Determine the probability of occurrence
f. Assess current control processes in place
g. Evaluateability to detect the failure mode
h. Assign a risk rating (A x B x C) relative to:

A: Severity of Impact - On a scale of 1 to 10, rateseriousness of the effect from the failure mode with 10 as catastrophic and 1 no impact.
B: Probability of Occurrence - likelihood that a cause and failure mode will occur with 10 as failure is certain and 1 is highly unlikely.
C: Ability to Detect - Rating your ability to detectfailure mode before putting the product into production or delivering it tocustomer. A rating of 10 indicates that you cannot detect the failure and 1 is where you have good controls in place to pickup the failure.

i. Take corrective actions on those failure modes with high risk ratings. results of your FMEA can be summarized on a worksheet.

3. Poka Yoke: Mistake proofing a product or service. Errors lead to defects and if you can catch the errors earlier, then you reducedefects. Certain work conditions tend to introduce errors: Adjustments, Infrequent Activities, Rapid Repetition Involved, and High Volume Loads with Compressed Time Frames. Once you’ve identified the error prone conditions, drill down toroot causes and see if you can design an error proof way of doingwork.

"A very few American companies are counted amongworld-class leaders in quality management. But thousands upon thousands of other companies have yet to take that all important first step to ensure their products and services deliver to each customer a dependable high level of quality. American economy will either fully integrate itself into new and evolving global markets, or large parts of it are likely to be left behind as foreign competitors absorb greater and greater shares ofonly market that really matters anymore: global market."
- Quality in America: How to Implement a Competitive Quality Program by V. Daniel Hunt