Lean Six Sigma

DMAIC process

Benefits
• Uncover the deep root causes to a problem.
• Require teams to make decisions based on data, whenever
possible.
• Sustain and leverage results going into the future.

DEFINE: Define the Project
• Define the project’s purpose and scope and get background on the process and
customer.
MEASURE: Measure the Current Situation
• Focus the improvement effort by gathering information on the current situation.
ANALYZE: Analyze to Identify Causes
• Identify root causes and confirm them with data.
IMPROVE: Implement Solutions and Evaluate Results
• Develop, try out, and implement solutions that address root causes.
• Use data to evaluate both the solutions and the plans used to carry them out.
CONTROL: Standardize and Make Future Plans
• Maintain the gains by standardizing work methods or processes.
• Anticipate future improvements and preserve the lessons from this effort.

DEFINE phase

• Understand the Voice of the Customer


• Develop charter


• Map the process

To fully define the scope and purpose of your project, you will need to understand the
boundaries of the process you’re trying to improve and the needs of the customers of that
process. This information will be captured in your team charter, which will also include
other information such as expected resource needs and projected timeline.
In practice, there is usually some back-and-forth between these activities as you work to
define a project that is both important and doable.

SIPOC Map

The SIPOC map allows us to determine how the process links
with suppliers, customers, and other processes by identifying
• Where the inputs to the process come from
• Who the recipients of the outputs are

The SIPOC Model Process Map is used to:
• Understand a high level view of the process.
• Understand how this process links with suppliers, customers, and other processes.
• Communicate the high level process to others.
• Manage interdependent processes particularly when tied to measurement.
The components of the SIPOC Model include:
• Suppliers: Who supplies the inputs to the process.
• Inputs: Information/suppliers/material needed to perform the work.
• Process: Steps (activities) needed to transform the inputs to the outputs.
• Outputs: Transformed (value-added) information/products transmitted to the
customer.
• Customers: Internal or external customers who use the outputs and benefit directly
from the process.

• A high-level map usually has between 3 and 7 steps
• Each step has a verb which describes the activity and a noun
which describes what is flowing through the process

With Top Level Process Maps, the focus is on main steps, not details. Here, you are NOT
concerned with loops or errors.
• List 5–7 key steps that capture the main action that takes place in the process.
• Optionally, add substeps below each major step.

Measure Phase

Process Map

• A process map shows the sequence of steps in a process and
who carries out each step
• It highlights customer-supplier relationships within the process
• Time flows down the chart—the first step is at the top, the last at the bottom
• People involved are listed across the top of the chart
• Each step is placed in the column of the person who carries it out
• Arrows show relationships between steps
• Different symbols may be used for different types of activities (decisions, actions, meetings…)

Value Stream Analysis
• One possible next step is to analyze the process to identify waste in the process
• After waste is identified, it can be removed
• This makes the process more efficient while maintaining quality
• Non-value-added activities within a process may represent potential waste

Data Help Us...
• Separate what we think is happening from what is really happening
• Confirm or disprove preconceived ideas and theories
• Establish a baseline of performance
• See the history of the problem over time
• Measure the impact of changes on a process
• Identify and understand relationships that might help
explain variation
• Control a process (monitor process performance)
• Avoid “solutions” that don’t solve the real problem

Note: Most people are accustomed to seeing results measures—data that is taken on processoutput. By the time you collect results data, however, it’s usually way too late to fix theprocess. A more effective improvement strategy is to develop process measures takenon the process while it is operating. Process measures can give you real-time feedbackabout how well the process is operating, and allow you to detect problems sooner.

Data Collection
Make sure that
• The data you collect will give you the answers you need
• There is enough data on which to base reliable conclusions
• The full range of actual conditions is seen in the data
• Data isn’t “counted” twice
• The data collection form is easy to use and understand
• There is space for comments on the form
• You include only the information you will use
• You pilot the form on a small scale before using it widely
• Operational definitions are developed for the data you want to collect
• Everyone measures the same thing the same way, day after day
Total Variation = Variation in the Process + Variation of the Measurement System

Operational Definitions
• An operational definition is a precise description that tells how to get a value for the characteristic you are trying to measure
• It includes what something is and how to measure it
• An operational definition:
• Removes ambiguity so that all people involved have the same understanding of the characteristic or feature in question
• Describes your way of measuring that characteristic or feature

Features of Operational Definitions
• Must be specific and concrete
• Must be measurable
• Must be useful to both you and your customer
• There is no single right answer

• When an operational definition is specific and concrete, different people can use the definition and know that their data will be measured in the same way.
• Being measurable means you know how to assign a value (either number or yes/no) to a data point.
• To be useful, an operational definition must be meaningful to both you and your customers. That is, it should relate to how the customer will judge quality, and should allow a go/no go decision—“yes, we’ve met the customer’s need” or “no, we haven’t met the customer’s need.”
• There is no single right way to write an operational definition, but everyone who is measuring something must agree on the definition.

Key Points About Operational Definitions
• There is no single right answer for how to operationally define a measure
• The more specific, the better
• Plan on refining the definition after you try it out
• Training will help data collectors consistently apply operational definitions

Focus on the Variation
• When analyzing time-ordered data, you need to look at the variation, how the data values change from point to point
• On a time plot, the changes from point to point is referred to as the “range.”
• Certain patterns in the variation can provide clues about the source of process problems

What Is Variation?

• No two anythings are exactly alike
• How a process is done will vary from day to day
• Measurements or counts collected on process output will vary over time
• Quantifying the amount of variation in a process is a critical step towards improvement
• Understanding what causes that variation helps us decide what kinds of actions are most likely to lead to lasting improvement

Types of Variation
Special cause variation
• Temporary or local; specific.
• May come and go sporadically.
• Evidence of the lack of statistical control is a signal that a special cause is likely to have occurred.
• A process with special cause variation is called unstable.
What are some exa ples of special cause variation in your work?

Common cause variation
• Common to all occasions and places.
• Degree of presence varies.
• Each cause contributes a small effect to the variation in results.
• Variation due to common causes will almost always give results that are in statistical control.
• A process with only common cause variation is called stable.

Goal: Minimize Variation
• There will always be some variation in a process
• But we can work to minimize variation around a target

Special Cause Strategy
Goal is to eliminate the specific special causes; that is, to make an unstable process stable
• Get timely data so special causes are signaled quickly
• Take immediate action to remedy any damage
• Immediately search for a cause
• Find out what was different on that occasion
• Isolate the deepest cause you can affect
• Develop a longer-term remedy that will prevent that special cause from recurring or, if results are good, retain that lesson

Use early warning indicators, or “leading measures,” throughout your operation. Take data at the early process stages so you can tell as soon as possible when something has changed.
You may not need to complete DMAIC to address a special cause.
• Take remedial action: get this project back on schedule, help this angry customer,isolate the roduct already produced.
• See what changed at the point in time when special cause appeared. What was different then?
• It the cause is not clear, continue on and analyze causes.
• If the cause is clear, confirm it with additional data, if possible. Then develop longer-term action to prevent the special cause (if the impact was bad) or preserve it (if the impact was good).

Common Cause Strategy
• A process with only common causes is said to be “statistically stable” and “in statistical control”
• Merely being in statistical control does not mean the results of a system are acceptable
• Leaving the process alone is not improvement
• A different approach is needed to improve a stable system

Improving a Stable Process
• Common causes of variation can hardly ever be reduced by attempts to explain the difference between individual data
points if the process is in statistical control
• All the data are relevant
• When dealing with special causes, you focus on a few data points
• For common cause variation, you need to look at all the data points to fully understand the pattern
• Break all the data into categories by looking at when the problem occurs and when it doesn’t; where the problem occurs and where it doesn’t, etc.—this will help narrow the focus
• Processes in statistical control usually require fundamental changes for improvement

• When improving a stable system you don’t single out one or two data points. You need to look at all the data—not just high points or low points—not just the points you don’t like—not just the latest point.
• Improving a stable process is more complex than identifying a special cause. More time and resources are generally needed in the discovery process.
• When dealing with special causes, you focus on a few data points.
• For common cause variation, you need to look at all the data points to fully understand the pattern.

Control Charts

• Time ordered plot of results (just like time plots)
• Statistically determined control limits are drawn on the plot

• Based on the actual variation in data, we can determine how much variation is typical for a process. This spread of expected variation is indicated by lines called control limits.
• Statistical control limits are not based on what we would like the process to do.
They are based on what the process is capable of doing (= process capability).
They are computed from the data using statistical formulas.

Control Chart Uses
• Track performance over time
• Evaluate progress after process changes/improvements
• Statistical control limits establish process capability
• Statistical control limits are one way to separate common cause and special cause variation:
• Points outside statistical limits signal a special cause
• Can be used for almost any type of data collected over time

Constructing a Control Chart
1. Plot data in time order
2. Calculate the average.
• Call this average X (“X-bar”)
3. Determine the average of the ranges
• Call this average R (“R-bar”)
4. Multiply the average range by 2.66
5. Add that result to the average to get the Upper Control Limit (UCL)
• Subtract the result from the average to get the Lower Control Limit (LCL)
6. Draw the centerline and control limits on the chart Centerline = X (average of data)
UCL = X + 2.66 R
LCL = X – 2.66 R

1. Other formulas can be used to calculate the control limits, but the ones given here
have proven very reliable even when we have data with special causes.
2. The centerline is often drawn as a solid red line on a control chart, and the control
limits are drawn as dashed green lines. Those techniques make it easier to read the
charts.

Specification Limits vs. Control Limits
• Specification limits
• Set by the customer, management, or engineering requirements
• Describe what you want a process to achieve
• Control limits
• Calculated from the data
• Describe what the process is capable of achieving

What to Do Next?
Search for patterns in the data
• The team does not want the amount of damage to remain the same
• How can the data help them decide what to do next?
• When there are no special patterns in the data, you can get a better understanding by breaking the data into subgroups

What Is a Pareto Chart?
A Pareto chart is a graphical tool that helps you break a big problem down into its parts and identify which parts are the most important.
• Categories are across the bottom; number of instances is on the vertical axis.
• Height of bar represents relative importance of that aspect of the problem.
• Bars are arranged in descending order from left to right, except for the “other” bar which is always on the right.
• Height of vertical axis represents sum of all occurrences (not just the height of the tallest bar).

When to Use a Pareto Chart
Use a Pareto chart when:
• The problem under study can be broken down into categories
• You want to identify the “vital few” categories
• Number of occurrences can be counted for each category

Interpreting a Pareto Chart
Check to see if:
• The Pareto principle applies
• Check to see that a few categories account for most of the problems
• The “other” category is small
• Make sure the “Other” bar is not the tallest category
• If it is, the occurrences categorized as “other” need to be re-examined to see if new categories can be made
• Other ways to categorize the problem were considered
• Are there other ways to divide the problem into categories?
• Can the Pareto chart be redrawn to account for the cost or impact of the problem?

Reacting to Pareto Charts
If the Pareto principle holds, and a few categories are responsible for most of the problems:
• If any of the bars point to problems with simple solutions, by all means attack them, even if these problems are not the biggest bars
• Begin work on the largest bar; make another Pareto chart of the components of the largest bar
• Move to Step 3: Analyze to identify and confirm the causes of the specific problem
• Once you have worked to solve a piece of the problem, collect more data and create a new Pareto chart; compare the initial and the new Pareto side-by-side to see how your improvements have affected the problem

When the Pareto principle does not hold:
• If many categories are needed to account for most of the problem.
• If all bars are roughly the same height.
What to do:
• Look for another way to break down or categorize the problem.
• Adjust the counts for their impact or dollar cost.
• Normalize the data—adjust the counts by the opportunity for occurrences in their category. In the example above, different numbers of hours were worked in each department. Some departments had more opportunity for accidents than others. To see which department is more dangerous, it is better to turn the counts into a rate or percentage.

Cause-and-Effect Diagram

The cause-and-effect graphically displays people’s opinions about possible causes of a problem. It is also known as:
• A fishbone diagram because of its appearance.
• An Ishikawa diagram after its inventor.

Cause-and-Effect Guidelines

• Capture cause-and-effect relationships between uits and sub-units
• Do not use this tool as n alternative form o outlining
• Potential causes on the diagram must be erified with data to confirm that they are real causes
• Problem must be focused and specific before using this tool
• Broad problems lead to diagrams that have too many items
• They are tedious to construct, time-consuming, and very dfficult to verify with data

Create a cause-and-effect diagram
1. Review the focused problem (the head)
• Write the description of the problem
• Draw a box around it; then draw a spine that leads into the box
2. Identify possible causes
• Write possible causes on self-stick note
• Put one cause per self-stick note
3. Sort possible causes into clusters
• Post the notes on flipchart paper
4. Choose a cluster and label a main bone
• Select a cluster of notes
• Decide how to word the main bone or main cause
• Draw a main bone leading into the spine
• Write the label at the end of the main bone
5. Develop and arrange bones for that cluster
• Ask “why might this happen?” to develop medium and small bones
• Use ideas from the original cluster when they fit
• Write new ideas on notes and add them to the diagram
• If a cause belongs in two places, put in both places
• Check the cause-effect-logic periodically
• Add lines to logically link all causes
6. Develop other main bones
7. Select possible causes to verify data

Improve Phase

Verifying Causes
• A lot of thinking and effort goes into constructing a cause-and-effect diagram, tree diagram, or relations diagram
• But those diagrams only identify potential causes
• You need to collect data to confirm which potential causes actually contribute to the problem
• It is often tempting to think that because we have used a tool like a cause-andeffect diagram or a relationships diagram, we now know the answer. We have identified the causes and written them down. Now we can get on with solving the problem.
• In our rush to find solutions we forget that we have only identified potential causes. Our ideas and theories are not data, but they provide us with a place to start collecting data. Only when we have verified which potential causes are actual causes are we ready to move on to Step 4: Solutions.

Go to the Workplace
• A primary source of data on any problem is direct observation in the workplace
• The younger, less senior forklift drivers did not reduce their corner gouges after training—what else could account for their poor performance?
• Go into the workplace and observe

Neutral Observation
• Neutral observation includes specific descriptions of what you see or hear
• Neutral observation does not include:
• Judgments
• Values
• Advice
• Suggestions
• Interpretations
• Your story about “what it means”

Control Phase

Standardization Action Taken
Memo sent to managers at the seven other warehouses in the
company. Advised them to check for outmoded forklift adaptations
and refit them if appropriate. Also suggested they set up a system
to monitor the use of special equipment to avoid using modified
equipment when it is no longer required or appropriate. Included a
copy of this Improvement Story for documentation.

Maintaining the Gains

monitor how well the gains are maintained, the team plotted the number of
damaged cartons on a daily basis. This plot shows that the reduction in carton
damage is maintained several weeks after the clamp replacement.

Strategy for Continuation
1. Identify causes of remaining Type F damage in Production to Warehouse and Warehouse to Shipping. (Review causeand-effect diagram for other possible causes.)
2. Begin study of Type H damage (tip crush).
3. Remember to record all data in time order.