Introduction to Statistical Process Control SPC 2001 General

Introduction to Statistical Process Control (SPC) ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Process = Hose Four Important Properties: (1) Centering (2) Spread (3) Shape (4) Stability Over Time 10. 5 10 9. 5 Y axis = Weight (lbs) ã 2001, General Electric Company, USA. All rights reserved GE Confidential

5 ft 4 ft 3 ft 2 ft 1 ft 0 ft ã 2001, General Electric Company, USA. All rights reserved GE Confidential

5 ft 4 ft 3 ft 2 ft 1 ft 0 ft ã 2001, General Electric Company, USA. All rights reserved GE Confidential

What has changed? 5 ft Centering? ______ Spread? ______ 4 ft 3 ft 2 ft 1 ft 0 ft m Ti 4. 0 Time e 3. 0 2. 0 ã 2001, General Electric Company, USA. All rights reserved GE Confidential

5 ft 4 ft 3 ft 2 ft 1 ft 0 ft Time What has changed? Centering? ______ Spread? ______ ã 2001, General Electric Company, USA. All rights reserved GE Confidential

5 ft 4 ft 3 ft 2 ft 1 ft 0 ft Time What has changed? Centering? ______ Spread? ______ ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Avg Spread Stable? Yes/No ã 2001, General Electric Company, USA. All rights reserved GE Confidential

DMAIC = Doctor’s Detective-work - Hotel Detective-Work (Tempting to jump to conclusions/solutions? ) Analyze Phase = Diagnosis: Y= f(X, X, . . . , X) What Data To Track Performance (Y’s)? - Hospital Transactions (Where to shove the “Thermometers” to get data? ) Strategy #1 Detective-Work for an Unstable Process - Tracking the Y: Behavior Change = Learning Opportunity - Communicate SPC Effectively: Intuitive-Visual Approach Next Strategy #2 - Get the Right “Meaning” From Your Data (Signal VS Noise) - Bubba: On the Straight and Narrow Path? - Inventory Management - The Goal Detective-Work for a Stable Process - Tracking the Y and X’s Simultaneously: Relationships - Active vs Passive Detective-Work - The Multi-Vari Study: Mechanics Example ã 2001, General Electric Company, USA. All rights reserved GE Confidential

The Scene: • Bubba is known as “The straightest driver in Crick County”. • One evening at Charlie’s Bar, he boasted, ”I drive so straight, even a quart of Moonshine won’t throw me off target” • Fed-up, Junior told him to “put his money where his mouth is” -- a $100 bet. • They selected Charlie, the bar owner, to be the impartial gum dropper. ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Charlie: The Impartial Mediator Warm, sticky, pre-chewed gum arlie’s Ch Pre-Chewed G Bar and ã 2001, General Electric Company, USA. All rights reserved GE Confidential rill

You make the call: Did Bubba stay straight or loose $100? Junior says, “My gosh, it looks like you were swerving all over the place right from the beginning!” “Where’s my $100? ” ã 2001, General Electric Company, USA. All rights reserved GE Confidential

CONSIDER THIS. . . Warm, sticky, pre-chewed gum arlie’s Ch Pre-Chewed G Bar and rill Why do the pieces of gum not land at the exact same location? . . . even when Charlie is trying his best! ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Sources of Variation - Natural “Aim-to-Aim Variation” (due to Charlie) To develop common terminology, for now, let’s simply call this “Aim-to-Aim Variation” ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Now what do you think: Did Bubba stay straight or loose $100? Wouldn’t it be helpful to have some way of separating the steering variation from the natural, inherent “Aim-to-Aim” Variation? ã 2001, General Electric Company, USA. All rights reserved GE Confidential

The Control Chart Approach Sources of Variation: - Natural Aim-to-Aim Variation We can strategically use this natural, unavoidable variation (i. e. , the spread width) as a tool for detecting shifts in the truck’s centering. ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Like using a “Squelch” to separate Signals from the Noise: First, select the spread that we will declare as the “Baseline Behavior”. Then, determine boundaries or limits that should contain virtually all (say 99. 7%) of the “wiggling” Here, we will select “Natural Aim Variation”. (that is, if Charlie’s Aim continues to be the only Source of Variation present). (Setting the squelch threshold level) Suppose that whenever any point lands outside these “control limits”, an alarm will sound. (Detecting a signal being that’s above and beyond the squelch threshold) ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Sources of Variation - Aim Variation (only) - Aim Variation - Variation Due to Steering So, if any shifts in the centering occur, we will know! ã 2001, General Electric Company, USA. All rights reserved GE Confidential

That’s how control limits separate signals from noise UCL LCL No evidence of any changes in the true “centering” of the truck. All the “wiggling” of the data points was simply due to noise. ã 2001, General Electric Company, USA. All rights reserved GE Confidential

DMAIC = Doctor’s Detective-work - Hotel Detective-Work (Tempting to jump to conclusions/solutions? ) Analyze Phase = Diagnosis: Y= f(X, X, . . . , X) What Data To Track Performance (Y’s)? - Hospital Transactions (Where to shove the “Thermometers” to get data? ) Strategy #1 Detective-Work for an Unstable Process - Tracking the Y: Behavior Change = Learning Opportunity - Communicate SPC Effectively: Intuitive-Visual Approach Next Strategy #2 - Get the Right “Meaning” From Your Data (Signal VS Noise) - Bubba: On the Straight and Narrow Path? - Inventory Management - The Goal Detective-Work for a Stable Process - Tracking the Y and X’s Simultaneously: Relationships - Active vs Passive Detective-Work - The Multi-Vari Study: Mechanics Example ã 2001, General Electric Company, USA. All rights reserved GE Confidential

• April 1994: Manager pleased to see the in-process inventory drop to 15. In-Process Inventory (X) • Manager presents an award to Dept. 50 in honor of this achievement. 30 • Ceremony in the cafeteria: pizza and refreshments for all! • “Everyone should be proud of what you’ve accomplished”. 20 10 J F M A 1994 1995 Award given ã 2001, General Electric Company, USA. All rights reserved Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

• July 1994: Three consecutive months of inventory increases. • Manager wishes he could take back the award. In-Process Inventory (X) • “Recognition seems to have been backfired. ” 30 • Instead of holding the gains, Dept. 50 seems to have gotten complacent -- allowing inventory to creep back to where it was previously. • Manager decides: “This Mr. Nice Guy stuff doesn’t work” 20 10 J F M A M J 1994 ã 2001, General Electric Company, USA. All rights reserved J 1995 Manager regrets giving award Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

• Nov. 1994: In-process inventory rises to a value of 26! • Manager decides to kick anatomy and take names. In-Process Inventory (X) • He calls everyone in and demands they do something to keep inventories down. 30 • Following this tantrum, everyone in Dept. 50 treads lightly. Short of hiding material in the dark corners of the plant, they don’t know what to do. • So, they hold their breath and hope inventories drop. 20 10 J F M A M J 1994 ã 2001, General Electric Company, USA. All rights reserved J A S O N D 1995 No more Mr. Nice Guy Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

• June 1995: Manager has seen reduced levels of inventory since the end of last year. “Things are looking-up!” (Although nothing had been done to change the system) In-Process Inventory (X) • His learning: “Tough management style gets results!” Manager concludes: “Tough management style gets results!” 30 20 10 J F M A M J 1994 ã 2001, General Electric Company, USA. All rights reserved J A S O N D J F M A M J 1995 Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

“But I always made sure my decisions/actions were based on data! What could be wrong with that? ” • His decisions stemmed from interpreting the high and low points as signals. • However, as the control chart below clearly reveals, NONE OF THESE INDIVIDUAL POINTS WERE SIGNALS! In-Process Inventory (X) • They all reflect the same underlying system -- each should be interpreted alike. 30 UCL 20 LCL 10 J F M A M J 1994 ã 2001, General Electric Company, USA. All rights reserved J A S O N D J F M M J J A S O 1995 Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

In-Process Inventory (X) Manager concludes, “Tough management gets results!” 30 UCL 20 LCL 10 J F M A M J J A S O N D 1994 Award given J F M M J J A S O 1995 Manager regrets giving award No more Mr. Nice Guy • No matter how logical and persuasive the interpretation, and no matter how “right” you’ve been in the past -- noise is still noise. • Trying to interpret noise is nothing more than an exercise in wishful thinking. • The result -- pure fiction with no connection to reality. ã 2001, General Electric Company, USA. All rights reserved Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

“Failure to use control charts to analyze data is one of the best ways known to mankind to increase costs, waste effort, and lower morale. ” “It’s not what we don’t know that hurts. . . it’s what we know that ain’t so. ” - Dr. Donald J. Wheeler - Will Rogers False lessons cost us -- better to know how to use data wisely! In-Process Inventory (X) Manager concludes, “Tough management gets results!” 30 UCL 20 LCL 10 J F M A M J J A S O N D 1994 Award given ã 2001, General Electric Company, USA. All rights reserved J F M M J J A S O 1995 Manager regrets giving award No more Mr. Nice Guy Derived from Understanding Variation: The Key To Managing Chaos , Donald J. Wheeler, SPC Press. 1993. GE Confidential

Video Segment (Optional) Watch how one manager’s decisions and actions are affected when he incorrectly interprets. . . NOISE (typical, inherent, minute-to-minute process variation behavior) as if it were a SIGNAL (clearly different behavior -- above and beyond the usual noise) Do you know any operators, co-workers, or managers who have managed processes this way? ã 2001, General Electric Company, USA. All rights reserved GE Confidential

“Teen Smoking Turns Upward” - USA Today, June, 21, 1994 Percent By Year 19. 5 1993 19. 0% 1992 17. 3% 19. 0 18. 5 18. 0 17. 5 17. 0 1992 1993 Year ã 2001, General Electric Company, USA. All rights reserved GE Confidential

1984 18. 8 1985 19. 6 1986 18. 7 1987 18. 6 1988 18. 1 1989 18. 9 1990 19. 2 1991 18. 2 1992 17. 3 1993 19. 0 I and MR Chart for %Smokers A Signal Above the Noise? 21 Individuals %Smokers UCL=20. 71 20 19 18 X=18. 64 17 LCL=16. 57 16 84 Year Moving Range Year 85 86 87 88 89 90 91 92 93 3 UCL=2. 541 2 1 R=0. 7778 0 LCL=None ã 2001, General Electric Company, USA. All rights reserved GE Confidential

So, what can we say about the percentage of teenagers who smoke daily? Just this. . . • There is no evidence that the percentage of teenagers who smoke has increased. • Neither is there any evidence that this percentage has decreased during the past 10 years. • The only headline for these data that has any integrity is "No change in teen use of tobacco. " Anything else is just propaganda. So how do you avoid being persuaded by propaganda? • The beginning is to realize that while all data contain noise, only some data contain signals. • If you do not know how to separate the probable noise from the potential signals, you are susceptible to being misled by the noise in the data. • Others may use data to mislead you -- or you may even mislead yourself. Shewhart's charts are the simplest way to separate signals from noise. By the way, did you read the article about how the Trade Deficit soared last April? Oh well, that's another story -- or is it? ã 2001, General Electric Company, USA. All rights reserved GE Confidential

A Case Study. . . Selit Corp. ’s 4 th Quarter Sales Report. . . Time Same Quarter Last Year Last Quarter Current Quarter NEast SWest NCentral MAtlantic SCentral 1995_Q 4 952 1196 878 670 802 462 1996_Q 3 1996_Q 4 976 1148 1197 1337 688 806 743 702 807 781 447 359 (Values In Columns = Sales In Thousands) Open minitab worksheet file “Sales (Quarterly Data). mtw” ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Mid-Atlantic Same Quarter Last Year Last Quarter Current Quarter 802 807 781 Selit Corp. ’s 4 th Quarter Sales Report. . . “Kim, Mid-Atlantic sales were down 3. 2% in the fourth quarter and down 2. 6% from the previous year -- a downward trend! I’m very disappointed in your performance. You were once my best sales representative. I had high expectations for you. Now, I can only hope that our first quarter results show some sign of life. ” Kim felt her face get red. She knew she sold more units in 1996 than in 1995. “What does Halger know anyway, ” she thought to herself. “He’s just an empty suit. ” ã 2001, General Electric Company, USA. All rights reserved GE Confidential

South. West Same Quarter Last Year Last Quarter Current Quarter 1196 1197 1337 Selit Corp. ’s 4 th Quarter Sales Report. . . Hagler continued: “Terry, Southwest sales were super. You showed an 11. 7% increase in the fourth quarter and an 11. 8% increase over the previous year” -- An upward trend! Terry smiled. She wasn’t sure how she did so well, but she sure wasn’t going to change anything. ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Example: Process: Driving to work Measurement: Y = Travel time Brainstorming: List possible causes of both common and assignable causes Noise = “Common Cause” Variation Signal = “Assignable Cause” Variation (or “Special Cause”) The strategies for improving these two types of variation are very different !! More about this later. . . ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Example: Process: Monthly Telephone Bill Measurement: Y = $ Brainstorming: List possible causes of both common and assignable causes Noise = “Common Cause” Variation Signal = “Assignable Cause” Variation (or “Special Cause”) The strategies for improving these two types of variation are very different !! More about this later. . . ã 2001, General Electric Company, USA. All rights reserved GE Confidential

“Common Cause” Highway (Noise) Although we don’t know exactly where the next drop will land, we are reasonably confident it will fall somewhere on the “Common Cause” Highway. For this reason, a process that has only common cause variation is said to be. . . STABLE and PREDICTABLE. 30 20 10 “Special Cause” Highway (Signal) A process with Special Causes might better be represented by a highway through an earthquake zone: likely to suddenly shift to the side by several feet or more. For this reason, a process that has Special Causes of variation is said to be. . . UNSTABLE and UNPREDICTABLE. 30 20 10 ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Components of a Control Chart Y-axis (Response) UCL Upper Control Limit Central Line LCL Lower Control Limit TIME ã 2001, General Electric Company, USA. All rights reserved GE Confidential

In case you might dare to use the idea one day. . . When communicating statistical concepts, you could describe SPC as follows: "Tracking process behavior by way of sample measurements (plotted points)" can be thought of as being much like "tracking a rabbit by way of its droppings". Suppose the football-shaped droppings fall at regular 1 minute intervals, and when they hit the ground, they tend to roll around a bit (like random/sampling variation). So the game becomes, "How to use statistical tools as reliable indicators -- to keep us from being fooled by the "wobbling" (noise) that will always be present in the droppings (data) -- so we reliably detect whether the rabbit did or did not move to the left or right of the straight path. " Here are some rules/principles to help us know when to react. . . ã 2001, General Electric Company, USA. All rights reserved GE Confidential

How to Detect a Change in Process Behavior: (i. e. , When an X just made an impact of the process) (1) Point beyond UCL or LCL (+/- 3 Std Dev’s) (2) 7 consecutive points on the same side of the centerline (Minitab uses 9) (3) Trend: 6 consecutive points steadily increasing or decreasing (4) Repeated patterns/cycles We can ask Minitab to search for 8 out of control conditions automatically ã 2001, General Electric Company, USA. All rights reserved GE Confidential

ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Let’s see how we can plot 3 or 4 points…. X-Bar & Range Chart for Wheeler “Master” Data File: “Wheel_G. mtw” ã 2001, General Electric Company, USA. All rights reserved GE Confidential

ã 2001, General Electric Company, USA. All rights reserved GE Confidential

Here’s a blank Xbar&R chart (for future use) ã 2001, General Electric Company, USA. All rights reserved GE Confidential

This is how a Control Chart actually looks using Minitab… ã 2001, General Electric Company, USA. All rights reserved GE Confidential

How to think about SPC: A subtle but crucial distinction: • Control limits are not there to "allow us to produce stuff within specifications" -- they have nothing to do with tolerances in any way. • Control limits are there only to allow the process to scream at us when an instability has occurred, in order to get someone off his/her rear-end quickly enough to find-out "What just changed to cause that? " while the "tracks" remain fresh -- before the evidence disappears forever. • In this way, the limits help to prevent people from creating additional variation by tampering with a process that should have been left alone, and to keep people from wasting their time and effort looking around for "What just changed? " when there actually is no rational basis for being suspicious in the first place. • Control limits, when established in a rational manner, are an operational definition for "When to be suspicious". ã 2001, General Electric Company, USA. All rights reserved GE Confidential