SPC Press has released a 15-part seminar series taught by Donald Wheeler on their site, and is available on YouTube at https://www.youtube.com/watch?v=rP3i97Q6XjM&list=PLUeXRJ5a5UC1cDLxbZPs6Jg_d9m-nvIjO&ab_channel=DonaldJ.Wheeler.
Each session draws heavily from his book “Understanding Variation: The Key to Managing Chaos,” and lays the groundwork for understanding data analysis in a business context, highlighting the limitations of traditional data interpretation methods and introducing the fundamental principles of Process Behavior Charts (PBCs). He emphasizes using PBCs as a tool for gaining insight, making predictions, and driving meaningful process improvement.
Each video is about an hour in length, which will require 15 hours to go through (less if you increase the playback speed of the videos). To save you some time to be able to find specific content that you want to dig into, we have taken SPC Press’s overview and added more details.
Search Lean Six Sigma Books on Amazon >>>
Disclaimer: As an Amazon Associate, we earn commission from qualifying purchases.
Session 1 Review
Overview: This initial session of Dr. Wheeler’s seminar strongly advocates for a shift in how organizations understand and interpret data. It critiques the limitations of relying on simple comparisons and arbitrary specifications, introducing the power of Process Behavior Charts as a method for understanding variation, identifying true signals of change, and ultimately gaining the knowledge needed for meaningful process improvement and prediction. The core message is that by understanding and utilizing the “voice of the process” through PBCs, organizations can move beyond reactive management based on noise to proactive improvement based on knowledge.
- The Problem with Traditional Data Interpretation
- Data in Isolation is Meaningless: Dr. Wheeler asserts that data must be viewed within its context to be properly understood. Simple comparisons of single data points (e.g., month-to-month, year-over-year) are often limited and can lead to misleading conclusions. The first principle for understanding data is that “no data have meaning apart from their context.” Context informs interpretation and analysis.
- Limitations of Monthly Reports: Relying solely on monthly reports and simple comparisons is likened to “trying to drive a car by watching the yellow line in the rear view mirror.” These reports often lack the necessary context and obscure the underlying process behavior.
- Graphs are Essential for Understanding Data: Tables of numbers are difficult for the human brain to readily interpret. Time series charts and histograms are presented as the two basic and effective graphs for understanding data by placing it in context. The example of Babe Ruth’s and Roger Maris’s home run records illustrates how graphs provide a clearer understanding of performance consistency versus isolated peaks.
- Walter Shewhart’s Principles for Presenting Data
- Preserve Evidence for Predictions: Data should be presented in a way that allows for making valid predictions. This is the fundamental purpose of collecting and using data in a business setting.
- Summaries Should Not Mislead: Averages, ranges, and histograms used to summarize data should not lead users to take actions they wouldn’t take if they saw the data as a time series. This highlights the importance of considering the time-order sequence of data, which traditional statistical summaries often ignore.
- Time Series as the Basic Graph: Shewhart advocated for the time series chart as the primary tool for data analysis, supplemented occasionally by histograms.
- The Pitfalls of Specification-Based Management
- Arbitrary Numerical Targets: Management often sets arbitrary targets (plans, goals, budgets) for judging performance. This creates a “binary worldview” where performance is either “okay” or “in trouble.”
- On-Again, Off-Again Approach: This binary view leads to alternating periods of complacency and panic, hindering continual improvement.
- Three Ways to Meet Arbitrary Targets (often negative):
- Improve the system: Requires sustained effort and a focus on the process.
- Distort the system: Shifting blame and manipulating the process without real improvement.
- Distort the data: Manipulating the reported numbers.
- Three Ways to Meet Arbitrary Targets (often negative):
- Specifications vs. Process Voice: Specifications represent the “voice of the customer” and should not be confused with the “voice of the process.” Comparing data to specifications doesn’t provide insight into how the process works or how to improve it.
- The Limitations of Comparing Data to Averages
- Expected Variation: Values will naturally be above average about half the time and below average the other half. This simple comparison doesn’t indicate whether a change has occurred in the underlying process.
- Limited Comparison: Comparing a value to an average is a limited comparison that may not reveal meaningful changes.
- Introducing the Process Behavior Chart (PBC)
- Distinguishing Routine and Exceptional Variation: PBCs are designed to differentiate between the normal, “run-of-the-mill” variation inherent in any process (routine variation due to common causes) and variation that signals a change in the process (exceptional variation due to assignable causes).
- The Voice of the Process: The PBC is presented as the “voice of the process.” It uses data arranged in time order sequence, a central line, and control limits calculated from the data themselves.
- Interpreting Signals: Points falling outside the control limits are interpreted as signals of exceptional variation, indicating that a change has likely occurred in the process and warrants investigation.
- When to Fix It Chart: PBCs help determine “when to fix it” by highlighting when assignable causes are present. When only routine variation is observed, the process is predictable, and efforts should focus on fundamentally improving the system rather than reacting to individual data points.
- Basis for Prediction: When a process shows only routine variation, the past can be a reasonable guide to the future, allowing for predictions about process behavior.
- Three Types of Action
- Action on Outcomes (Specifications): Separating acceptable from unacceptable outcomes after the fact.
- Action on the Process (Process Behavior Charts): Identifying and addressing assignable causes of exceptional variation for future improvement. Refraining from reacting to routine variation.
- Action to Align Voices: Bringing the “voice of the process” (PBC limits) in line with the “voice of the customer” (specifications). This is difficult with an unpredictable process.
- Predictable vs. Unpredictable Processes
- Predictable Process: Exhibits only routine variation, with outcomes varying within natural process limits. The same common causes produce both good and bad outcomes. Improvement requires changing the system.
- Unpredictable Process: Displays both routine and exceptional variation due to assignable causes. PBCs will detect these signals, providing opportunities for learning and improvement by identifying the causes of these changes.
- Unpredictable processes vary without regard for natural limits or specifications.
- Examples Illustrating the Concepts
- Daily Percentage of Defective Pairs: A process operating predictably but at an unacceptable quality level. Highlights the difference between predictability and acceptability.
- In-Process Inventory: Demonstrates how traditional percent difference analysis can lead to unnecessary concern and wasted effort when the variation is routine.
- On-Time Shipments: Shows how small percent differences might mask signals of exceptional variation that are clearly visible on a PBC, leading to missed opportunities for improvement.
- Premium Freight: Illustrates how PBCs can retroactively reveal shifts and changes in a process that were missed by traditional analysis, and the importance of considering context when setting up and interpreting PBCs.
- Two Mistakes in Interpreting Data:
- Mistake 1: Treating Noise as Signal: Reacting to routine variation as if it indicates a meaningful change.
- Mistake 2: Failing to Detect a Signal: Missing actual changes in the process because the analysis method is inadequate.
- PBCs are designed to strike a balance between the economic consequences of these two mistakes by filtering out noise to reveal true signals.
- The Purpose of Analysis: Insight
- Data analysis should be driven by the goal of gaining insight into the process. Traditional monthly reports often lead to superficial explanations of noise rather than genuine understanding.
Session 2 Review
Overview: The session primarily focuses on the practical application of statistical process control (SPC) principles, emphasizing the importance of operational definitions, process behavior charts, and the dangers of misinterpreting routine variation as signals.
- The Crucial Role of Operational Definitions:
- He begins by highlighting the fundamental need for operational definitions as the bedrock of meaningful measurement and improvement. He draws upon Dr. W. Edward Deming’s framework, stating that an operational definition consists of three essential parts:
- A Criterion to be applied: “what do you want to accomplish” in a business context.
- A test of compliance: “by what method” will you assess if the criterion is met.
- A decision rule for interpreting the test results: “how will you know” if the goal has been achieved.
- He emphasizes that a simple specification, goal, or target is insufficient. An operational definition must include the method for achieving the goal and a clear way to determine when it has been met.
- Drawing on his experience with Dr. Deming, Wheeler notes that Deming consistently questioned goals by asking, “by what method?” and “how will you know?” Underscoring that without these answers, plans remain “just wishful thinking.”
- He begins by highlighting the fundamental need for operational definitions as the bedrock of meaningful measurement and improvement. He draws upon Dr. W. Edward Deming’s framework, stating that an operational definition consists of three essential parts:
- Process Behavior Charts as Operational Definitions of Improvement
- He asserts that process behavior charts themselves constitute an operational definition of how to get the most out of a process and achieve continual improvement. He aligns the components of a process behavior chart with the elements of an operational definition:
- What do you want to accomplish? The limits define the process’s potential. “The limits Define what a process can be made to do they approximate the ideal of what a process can achieve but it is operated up to its full potential.”
- By what method? The running record displays performance, and identified exceptional values guide investigation and improvement efforts.
- How will you know? By combining process potential and actual performance, you can judge how close the process is to its ideal.
- He asserts that process behavior charts themselves constitute an operational definition of how to get the most out of a process and achieve continual improvement. He aligns the components of a process behavior chart with the elements of an operational definition:
- The Importance of Simplicity in Analysis
- He advocates for simplicity in data analysis, quoting “keep it simple stupid still applies.” He warns against blindly using complex statistical techniques enabled by software without understanding them. He illustrates this with the example of weekly production data entered based on month-end paperwork habits, highlighting how easily underlying patterns can be obscured by the way data is collected and reported.
- The Pitfalls of Misinterpreting Data and Setting Inappropriate Goals
- The seminar provides several compelling examples of how focusing on targets and percent differences without understanding process behavior can lead to distortion of data and the system.
- The “first-time approval rate” example demonstrates how a target set by corporate led local teams to manipulate the testing process to achieve the desired quarterly average, masking underlying issues and even incurring unnecessary costs (“before they take the first test they take a preliminary test start counting to zero”). This highlights the principle that “give me a goal give me a Target I’ll figure out how to meet it distort the data distort the system whatever it takes.”
- The “pounds scrapped” example illustrates the confusion that can arise from relying solely on percentage comparisons without visualizing the data over time, especially in the presence of seasonal variation. “This suggests the process is doing better than average” versus “This suggests which process is doing slightly worse than before so which comparison’s right?” He emphasizes that “the monthly report hides Cycles like this” and that “if you have seasonality in your data at least you can see it on an XMR chart instead of sweeping it under the rod.”
- The “OSHA reportables” anecdote satirizes the futility and potential for manipulation when goals are set arbitrarily without understanding the underlying process. “If the executives could lower the accident rate by setting a goal then why didn’t they do that this year? If they cannot lower the accident rate by setting a goal what will be the effect of their setting a goal? Distort the data distort the system.” He stresses that “goal setting is often an act of desperation. It is no substitute for examining and learning about the underlying process.” Accidents should be viewed as an undesirable byproduct of operations, requiring study and mitigation efforts regardless of whether they exceed an arbitrary target.
- The Importance of Using the Right Data and Disaggregation
- He emphasizes that using the right data is crucial for meaningful analysis. The “on-time closings” example shows how aggregated count data can be misleading and fail to identify specific areas for improvement.
- He introduces the concept of “report card charts” and stresses the need to disaggregate (break down into groups or categories) data to understand the variation within individual components of a system.
- Setting goals on processes exhibiting uncontrolled variation is futile until the assignable cause of the unpredictability is identified.
- Recognizing the Impact of Changes Over Time with Process Behavior Charts
- The “Department 13” case study demonstrates how process behavior charts can be used to assess the real impact of deliberate changes to a process. By dividing the data into segments corresponding to the interventions and calculating separate limits, one can determine if the changes resulted in a statistically significant shift in process behavior.
- This example also highlights the importance of looking at multiple metrics to understand the overall effect of changes, as improvements in one area (material cost) can lead to deterioration in others (scrap rate in Department 14). This underscores the danger of optimizing parts of a system in isolation. “The system will not be optimized if you optimize each department separately.”
- Understanding Variation in Rare Events
- He addresses the challenge of analyzing rare events, such as spills. Traditional count charts (like c-charts) can be insensitive in these situations.
- He advocates for measuring the interval between events (e.g., days between spills) and using an XMR chart on these interval data to detect shifts in the rate of occurrence. This provides a more sensitive approach for identifying when the frequency of rare events is changing. “Instead of recording how many spills do we have this month we look at the date for each spill… how many days between spills.” This technique allows for signal detection even with limited data.
- Distinguishing Between Common Cause Variation (Noise) and Special Cause Variation (Signals)
- A central theme throughout the session is the critical distinction between routine variation (common cause or noise) and assignable cause variation (special cause or signals).
- The “pizza for lunch” anecdote brilliantly illustrates how managers often attribute outcomes to specific interventions (awards or reprimands) when the changes are simply due to the inherent variability of the process.
- He emphasizes that process behavior charts are essential for filtering out noise and identifying true signals that warrant investigation and action.
- He warns against the dangers of interpreting noise as signals. Attempts to explain random variation lead to “pure fiction” and can result in misguided actions.
- The Importance of a Shift in Thinking:
- He concludes by emphasizing that SPC is not just about applying tools but about adopting a new way of thinking about processes and data.
- He acknowledges the “Wall of Resistance” stemming from accumulated superstitious learning and the need for a “leap of faith” to truly understand and benefit from process behavior charts.
- Ultimately, he positions process behavior charts as essential for effective management, operations, and data analysis, enabling the separation of signals from noise, providing an operational definition of improvement, and offering a fundamental step in understanding data.
Session 3 Review
Coming soon
Session 4 Review
Coming soon
Session 5 Review
Coming soon
Session 6 Review
Coming soon
Session 7 Review
Coming soon
Session 8 Review
Coming soon
Session 9 Review
Coming soon
Session 10 Review
Coming soon
Session 11 Review
Coming soon
Session 12 Review
Coming soon
Session 13 Review
Coming soon
Session 14 Review
Coming soon
Session 15 Review
Coming soon