botchagalupe/ProfoundDeming · Datasets at Hugging Face

text stringlengths 0 1.44k
Continuing our example, let’s say his way resulted in less variation. The cannons’ mouths might vary between 83.5 mm and 84.5 mm. Less variation meant a “tighter” tolerance limit. The higher the cannon’s quality, the more effective its range and accuracy. The tighter a cannonball fit in a cannon’s mouth, the less air could get around the cannonball. The less air, the more explosive the force of the gunpowder on the cannonball, allowing the cannon to shoot farther.

Growing up poor as he did, I imagine frugality was almost in Ed’s DNA. The sheer amount of waste generated at Hawthorne Works must have boggled his mind. Surely, there was a way to improve this.

From his background in non-determinism, Ed understood that randomness and variation are simply facts of life . . . even in standardized manufacturing processes. He must have mulled for hours on how to find the solution to process deviations and defects inherent to the operations of Hawthorne.

As he came to find out, the answer lies with mathematics and statistics. Little did Deming know that he was about to get a front-row seat to the next turning point in the history of quality in the form of Dr. Walter Shewhart.

Deming’s Journey to Profound Knowledge - How Deming Helped Win a War, Altered the Face of Industry, and Holds the Key to Our Future - Part 1 - Chapter 4: The Root of All Evil

To understand Shewhart, we must first understand the theory of variation, and to understand variation, we must not only understand the history of quality but the history of measurement and probability as well.

Take, for instance, the time during the Peloponnesian War. Sparta besieged the city-state of Plataea. Finding themselves at an impasse, the Spartans built what amounted to a containment wall around the city. They stationed a few guards, and the rest went home. At some point, desperation, starvation, or deprivation would open the city gates; the Spartans simply needed to wait them out.

If necessity is the mother of invention, desperation is its muse. A soothsayer and a general hatched a bold plan: build ladders to scale the walls under the cover of night and escape between the enemy’s encampments. After making the decision, the only real question was: how tall did the ladders need to be?

Since the Plataeans couldn’t exactly climb to the top of the walls and use a tape measure in one hand while fending off Spartans with the other, they devised another means. They counted the bricks from afar.

According to Thucydides’ History of the Peloponnesian War, one side of the Spartans’ wall was. . . facing the town, at a place where the wall had accidentally not been plastered. A great many counted at once, and, although some might make mistakes, the calculation would be oftener right than wrong; for they repeated the process again and again. . . . In this manner they ascertained the proper length of the ladders . . . It was a narrow escape, but they pulled it off.

Saved by standardization. The bricks were all made of a common—a.k.a. standard-ish—size. All they needed to do was count how many brick layers there were and multiply that by the height of an everyday brick. Humans being human, though, not everyone agreed on how many layers of bricks there were. So, they had a lot of people count. Out of ten people, let’s say that six counted forty layers and two of them counted forty-one. Guy number nine was dead sure there were only thirty-nine layers. And the tenth guy reported fifty-nine, though he sort of slurred his words as he did.

What number should the ladder makers use? Thirty-nine? Forty? Forty-one? Fifty-nine? All of them couldn’t be right, of course; three must be in error. The Plataeans could have sent out even more people to count the layers, but at some point, the Spartans would start getting suspicious. They’d have to use the numbers they had. If they built the ladders too long, it’d mean more men to carry each one, making it harder to be stealthy. If they built them too short, well, what was the point? There was no certainty. It came down to a question of confidence: how sure were the would-be escapees in each of the four numbers?

They had zero confidence in the guy who counted fifty-nine bricks high. Nobody else counted anything near his number, and it sounded like he’d found the last bottle of wine in town, anyway. Thirty-nine was in the ballpark, but only one person got that number. In the realm of “how good is good enough,” they’d much rather err on the side of caution. The two people saying it was forty-one were known to have excellent eyesight, and most of the others counted forty layers, anyway. Everyone could feel fairly confident that the wall was somewhere around forty to forty-one bricks high. Whichever number they settled on, their estimates were good enough to get them over the wall.

In the end, out of the two hundred and twenty who made the attempt, two hundred and twelve escaped. This is statistics, variance, and probability in a nutshell (obviously, it’s more complicated than that). It’s not about certainty; that’s determinism. One of the linchpins of Deming’s System of Profound Knowledge is understanding uncertainty; that is, applying statistics to variation (which in this case would be the differences between how many bricks each person counted). This allows us to quantify certainty versus uncertainty.

Put another way, statistics is about how confident you feel when dealing with uncertainty. It’s about how probable an outcome is. If you let go of an apple, there is an extremely high probability it will hit the floor. If you pick any American, there’s a 50% chance they earn below the median income. If you pick a spot on the globe at random, there’s a 71% chance you’ll hit water. It’s all about chance and probabilities.

Deming gives us a great, simple example of variability. If you ask three people to count the number of people in the room, you might get three different answers. The answers depend on each counter’s definition of “the room.” Should the count include the people serving food or be limited to the guests? Should it include the open patio attached to the room?

Nobel Prize–winning physicist Percy Williams Bridgman was also concerned with variation when creating synthetic diamonds using extreme pressures. His gauges kept breaking down when they worked under extreme pressure, so he had no idea what pressure levels he had reached. This work led him to describe a general philosophical doctrine called operationalism. It is based on the idea that we can know the meaning of something only if we have a way of measuring it.
In 1927, Bridgman published The Logic of Modern Physics, examining how scientists define measurements. This work later inspired Shewhart’s and Deming’s ideas around what Bridgman coined an operational definition. Deming defined an operational definition in his book The New Economics as a procedure agreed upon to translate a concept into a precise measurement.

Operational definition became a key component of Deming’s theories of knowledge and variation.

While Ed was interning at Hawthorne Works, a man at the research arm of Western Electric was grappling with how to minimize manufacturing defects. Dr. Walter Shewhart believed there had to be a more economical way than simply standardizing production and using go/no-go tolerance limits. He wanted to minimize the variation between each telephone made. Being a physicist, his instinct was to try to solve the problem mathematically. Whereas basic statistics uses overall averages, he used a form of statistics that specifically averages the variation of defects—known as standard deviation. While standard deviation had been historically used in some areas of science, particularly with non-determinism, he was the first to apply non-deterministic methods to manufacturing. Curiously enough, standard deviation came about, in part, from counting
stars.

The Celestial Police

At its core, astronomy has always been about measuring. When do certain planetary bodies appear and disappear? What’s the distance from the horizon to this constellation, and how does that change over the four seasons? How long is daylight throughout the year? For that matter, what’s a year? Thousands of observations by amateurs and professionals alike, all from different parts of the world, all writing down everything by hand—can you imagine?

Astronomy began making great strides during the Scientific Revolution. In 1543, Copernicus published his observations supporting his theory that the solar system revolves around the sun (not the Earth).

In the early 1600s, Johannes Kepler wrote the laws of planetary motion. Based on his observations, Kepler noticed a pattern in planets’ sizes and movements across the sky. He spent years coming up with a mathematical formula that could explain why planets behaved the way they did. He wanted a perfect formula but would settle for whatever best fit the measurements he had. He had assumed that early astronomers had dismissed the idea of elliptical orbits; for years, he’d imagined that planets traveled in a perfect circle around the sun.

No matter how hard he squeezed, the data told him he was wrong. He tried an egg-shaped orbit, which didn’t work. Since circles and eggs weren’t the answer, he finally tried an ellipse. It fit. Not perfectly, mind you . . . but good enough.

Armed with this knowledge, he flipped things on their head. He had reverse-engineered the formula from the observations. Now that he had the formula, he used it to predict where other planets in the solar system would be—from observed measurements to a formula that best fit the data. From a best-fit formula to predicting data—statistics at its finest. Armed with this information, Kepler believed there should be a planet between Mars and Jupiter. He was wrong . . . but he was kind of right too.

One hundred and forty years after Kepler’s death, another astronomer used the additional century’s worth of data and developments to create his own predictive formula, dubbed the Titius-Bode law. It, too, predicted a planet between Mars and Jupiter, as well as predicting the distance of a planet beyond Saturn.

Lo and behold, in 1781 astronomers found a planet beyond Saturn: Uranus.

The Titius-Bode law gained credibility. Soon, astronomers were searching for the “lost” planet between Mars and Jupiter. After conversing with the discoverer of Uranus, a Hungarian astronomer assembled a crack team of twenty-four fellow astronomers. Their purpose: to coordinate their efforts to systematically search for it. He dubbed them the Celestial Police.

One of these astronomers was an Italian priest in Palermo named Giuseppe Piazzi. As fate would have it, he had already discovered the lost planet before his invitation to join the Celestial Police arrived in the mail. As irony would have it, he discovered it by accident.

That New Year’s Day in 1801, Piazzi didn’t know what he was looking at. It looked like a comet, but it acted like a planet. He made his observations for a little over a month before he became ill. By the time he’d sent notes to other astronomers who could observe the unidentified orbiting object, it had been obscured by the sun. By the end of the year, Piazzi knew it would have almost completed its orbit and should be visible again . . . but he couldn’t find it. He’d only been able to make twenty-four observations—not enough to predict where it would reappear.

A twenty-four-year-old German child prodigy decided to tackle the problem of how to use scant data to create a formula to predict where this thing should be. In three months, Carl Friedrich Gauss developed “Gauss’s method.” (With just a bit of tweaking, Gauss’s method would later become the basis for global positioning systems, or GPS.) Using Gauss’s method, the Hungarian chief of the Celestial Police spotted the dwarf planet Ceres, the largest object in our solar system’s main asteroid belt between Mars and Jupiter, just a few days shy of the one-year anniversary of Piazzi’s first observation.

Six years later, Gauss abandoned pure mathematics (he didn’t think it was worth it, apparently) and became an astronomer himself. While pursuing astronomical calculations (literally), Gauss laid the foundation for the statistics we’re talking about here.

This was Gauss's way of quantifying the randomness of his and other astronomers’ observations. With the Plataeans’ ladders, being a foot off was good enough. When dealing with pinpricks in the night sky that constantly change position, he needed to be a lot more precise. Gauss needed to remove as many measurement errors as possible. His method was essentially a way to quantify what the Plataeans did instinctively. He called these variations the mean error; the statistician Karl Pearson would come to refer to this as standard deviation—a core principle Shewhart used to create Statistical Process Control.

Getting Quality Under Statistical Process Control

It’s eerie how much of Walter Shewhart’s life echoes Deming’s (or maybe the other way around). He was born nine years before Deming on a rural farm in western Illinois. In 1910, he enrolled at the University of Illinois at Urbana-Champaign. He earned his doctorate in physics there in 1917.

He taught physics in Wisconsin, but that changed when Congress voted to enter World War I after the sinking of the Lusitania. The US Army, Navy, and budding Air Force placed $22 million worth of orders—about $500 million today—for communications equipment with Western Electric. Spurred by a sense of patriotism, Shewhart left academia to join the company’s engineering department. He moved his family to Brooklyn and commuted to his office at Western Electric’s building at 463 West Street overlooking the Hudson. Of course, the vast majority of Western Electric’s manufacturing took place at Hawthorne Works. Sometime between the end of the World War I and 1922, Shewhart became interested (some might say obsessed) with improving production quality. He was convinced that he could take the mathematical statistics concepts he’d learned as a physicist and apply them to production processes. On a Friday in May 1924, he summarized some of his ideas in a one-page memo to his boss, George Edwards.

As Isaac Asimov is attributed as saying, “The most exciting phrase to hear in science, the one that heralds new discoveries, is not ‘Eureka!’ (I found it!) but ‘That’s funny . . . ’” That’s how I feel about Shewhart’s memo. It should have been a eureka moment. Instead, there was no fanfare. No ticker tape parade.

Nobody knew the history of manufacturing had made one of its most important turning points.

Shewhart created a statistical system to improve production quality. He took the idea of tolerance limits and flipped it on its head with statistics. Instead of looking at telephones as good enough or not—that black-and-white deterministic thinking the world was used to—Shewhart quantified the variation.

He could now tell the managers at Hawthorne what percentage of products fell statistically within their tolerance limits. Far more importantly, it gave managers a method to track variation. If you can track variation, then you can trace variation to better understand why a production line creates defects and detects them much earlier in the process.

In essence, Shewhart created a way to continually improve any manufacturing line. This was the first time factory managers had been given a tool to let them manage the uncertainty in production.

His method, Statistical Process Control, let managers compare variation across workers and machines. The more managers could find and fix the causes behind factory defects, the more products they could produce that fell within their tolerance limits, thus improving their overall quality. And higher quality meant less waste; less waste allowed manufacturers to do more and more with less and less.

Think of an automaker doing a recall because of a braking issue. Tens of thousands of cars all have to be brought into their local dealerships to get their brakes fixed, costing the car company a fortune. In general, it’s cheaper to produce higher-quality brakes in the first place than to fix a mistake after the fact.

Shewhart’s genius was taking statistics from academia—physics, astronomy, biology, etc.—and applying it to Hawthorne’s assembly lines. He wanted to cut down on how much Western Electric wasted. For the first time in history, we had finally stepped beyond a simple go/no-go approach to making things.

Tracking cold, hard numbers allowed managers to track patterns in the data. Shewhart’s method was a paradigm shift. It was the first example of anti-Taylorism, where using math and statistics enabled management to see defects as results of the process instead of the workers. Before this, most managers viewed their employees like Stradivari did his apprentices. If the product was bad, it must be the workers’ fault. Stradivari never stopped to consider whether the woodcutter had sold him wood from a diseased tree. Never stopped to see if the apprentices’ tools were sufficient. Never once considered that he himself might be a poor teacher. His reaction was to blame the worker.

By 1929, Shewhart had formalized this new method of tracking and tracing variation. Basically, he applied the scientific method to manufacturing. Before this, manufacturing was a linear process. You figured out what you wanted and how many. You made them. Then you inspected and threw away the defects.
Shewhart turned this into a cycle, what Ed would later call the Shewhart Cycle: Figure out what you want, make it, inspect it, figure out what caused the defects, go fix it, and then go through the whole cycle again, using feedback from your mistakes to continuously improve production quality.

Even after Deming tweaked the Shewhart cycle, he still referred to it throughout his life as the Shewhart wheel. Despite this, his students in Japan called it the Deming cycle. Today, you might recognize it as the “plan, do, check, act” method, or simply the PDCA cycle.

In his later years, however, Deming came to rename check as study. To his way of thinking, check was too much like the go/no-go inspection process of checking manufactured products. He believed the better term was study, which implied approaching the results with a scientific curiosity to investigate and understand why things turned out the way they did.

Why Does It Vary?

Continuing our example, let’s say his way resulted in less variation. The cannons’ mouths might vary between 83.5 mm and 84.5 mm. Less variation meant a “tighter” tolerance limit. The higher the cannon’s quality, the more effective its range and accuracy. The tighter a cannonball fit in a cannon’s mouth, the less air could get around the cannonball. The less air, the more explosive the force of the gunpowder on the cannonball, allowing the cannon to shoot farther.

Growing up poor as he did, I imagine frugality was almost in Ed’s DNA. The sheer amount of waste generated at Hawthorne Works must have boggled his mind. Surely, there was a way to improve this.

From his background in non-determinism, Ed understood that randomness and variation are simply facts of life . . . even in standardized manufacturing processes. He must have mulled for hours on how to find the solution to process deviations and defects inherent to the operations of Hawthorne.

As he came to find out, the answer lies with mathematics and statistics. Little did Deming know that he was about to get a front-row seat to the next turning point in the history of quality in the form of Dr. Walter Shewhart.

Deming’s Journey to Profound Knowledge - How Deming Helped Win a War, Altered the Face of Industry, and Holds the Key to Our Future - Part 1 - Chapter 4: The Root of All Evil

To understand Shewhart, we must first understand the theory of variation, and to understand variation, we must not only understand the history of quality but the history of measurement and probability as well.

Take, for instance, the time during the Peloponnesian War. Sparta besieged the city-state of Plataea. Finding themselves at an impasse, the Spartans built what amounted to a containment wall around the city. They stationed a few guards, and the rest went home. At some point, desperation, starvation, or deprivation would open the city gates; the Spartans simply needed to wait them out.

If necessity is the mother of invention, desperation is its muse. A soothsayer and a general hatched a bold plan: build ladders to scale the walls under the cover of night and escape between the enemy’s encampments. After making the decision, the only real question was: how tall did the ladders need to be?

Since the Plataeans couldn’t exactly climb to the top of the walls and use a tape measure in one hand while fending off Spartans with the other, they devised another means. They counted the bricks from afar.

According to Thucydides’ History of the Peloponnesian War, one side of the Spartans’ wall was. . . facing the town, at a place where the wall had accidentally not been plastered. A great many counted at once, and, although some might make mistakes, the calculation would be oftener right than wrong; for they repeated the process again and again. . . . In this manner they ascertained the proper length of the ladders . . . It was a narrow escape, but they pulled it off.

Saved by standardization. The bricks were all made of a common—a.k.a. standard-ish—size. All they needed to do was count how many brick layers there were and multiply that by the height of an everyday brick. Humans being human, though, not everyone agreed on how many layers of bricks there were. So, they had a lot of people count. Out of ten people, let’s say that six counted forty layers and two of them counted forty-one. Guy number nine was dead sure there were only thirty-nine layers. And the tenth guy reported fifty-nine, though he sort of slurred his words as he did.

What number should the ladder makers use? Thirty-nine? Forty? Forty-one? Fifty-nine? All of them couldn’t be right, of course; three must be in error. The Plataeans could have sent out even more people to count the layers, but at some point, the Spartans would start getting suspicious. They’d have to use the numbers they had. If they built the ladders too long, it’d mean more men to carry each one, making it harder to be stealthy. If they built them too short, well, what was the point? There was no certainty. It came down to a question of confidence: how sure were the would-be escapees in each of the four numbers?

They had zero confidence in the guy who counted fifty-nine bricks high. Nobody else counted anything near his number, and it sounded like he’d found the last bottle of wine in town, anyway. Thirty-nine was in the ballpark, but only one person got that number. In the realm of “how good is good enough,” they’d much rather err on the side of caution. The two people saying it was forty-one were known to have excellent eyesight, and most of the others counted forty layers, anyway. Everyone could feel fairly confident that the wall was somewhere around forty to forty-one bricks high. Whichever number they settled on, their estimates were good enough to get them over the wall.

In the end, out of the two hundred and twenty who made the attempt, two hundred and twelve escaped. This is statistics, variance, and probability in a nutshell (obviously, it’s more complicated than that). It’s not about certainty; that’s determinism. One of the linchpins of Deming’s System of Profound Knowledge is understanding uncertainty; that is, applying statistics to variation (which in this case would be the differences between how many bricks each person counted). This allows us to quantify certainty versus uncertainty.

Put another way, statistics is about how confident you feel when dealing with uncertainty. It’s about how probable an outcome is. If you let go of an apple, there is an extremely high probability it will hit the floor. If you pick any American, there’s a 50% chance they earn below the median income. If you pick a spot on the globe at random, there’s a 71% chance you’ll hit water. It’s all about chance and probabilities.

Deming gives us a great, simple example of variability. If you ask three people to count the number of people in the room, you might get three different answers. The answers depend on each counter’s definition of “the room.” Should the count include the people serving food or be limited to the guests? Should it include the open patio attached to the room?

Nobel Prize–winning physicist Percy Williams Bridgman was also concerned with variation when creating synthetic diamonds using extreme pressures. His gauges kept breaking down when they worked under extreme pressure, so he had no idea what pressure levels he had reached. This work led him to describe a general philosophical doctrine called operationalism. It is based on the idea that we can know the meaning of something only if we have a way of measuring it.

In 1927, Bridgman published The Logic of Modern Physics, examining how scientists define measurements. This work later inspired Shewhart’s and Deming’s ideas around what Bridgman coined an operational definition. Deming defined an operational definition in his book The New Economics as a procedure agreed upon to translate a concept into a precise measurement.

Operational definition became a key component of Deming’s theories of knowledge and variation.

While Ed was interning at Hawthorne Works, a man at the research arm of Western Electric was grappling with how to minimize manufacturing defects. Dr. Walter Shewhart believed there had to be a more economical way than simply standardizing production and using go/no-go tolerance limits. He wanted to minimize the variation between each telephone made. Being a physicist, his instinct was to try to solve the problem mathematically. Whereas basic statistics uses overall averages, he used a form of statistics that specifically averages the variation of defects—known as standard deviation. While standard deviation had been historically used in some areas of science, particularly with non-determinism, he was the first to apply non-deterministic methods to manufacturing. Curiously enough, standard deviation came about, in part, from counting

stars.

The Celestial Police

At its core, astronomy has always been about measuring. When do certain planetary bodies appear and disappear? What’s the distance from the horizon to this constellation, and how does that change over the four seasons? How long is daylight throughout the year? For that matter, what’s a year? Thousands of observations by amateurs and professionals alike, all from different parts of the world, all writing down everything by hand—can you imagine?

Astronomy began making great strides during the Scientific Revolution. In 1543, Copernicus published his observations supporting his theory that the solar system revolves around the sun (not the Earth).

In the early 1600s, Johannes Kepler wrote the laws of planetary motion. Based on his observations, Kepler noticed a pattern in planets’ sizes and movements across the sky. He spent years coming up with a mathematical formula that could explain why planets behaved the way they did. He wanted a perfect formula but would settle for whatever best fit the measurements he had. He had assumed that early astronomers had dismissed the idea of elliptical orbits; for years, he’d imagined that planets traveled in a perfect circle around the sun.

No matter how hard he squeezed, the data told him he was wrong. He tried an egg-shaped orbit, which didn’t work. Since circles and eggs weren’t the answer, he finally tried an ellipse. It fit. Not perfectly, mind you . . . but good enough.

Armed with this knowledge, he flipped things on their head. He had reverse-engineered the formula from the observations. Now that he had the formula, he used it to predict where other planets in the solar system would be—from observed measurements to a formula that best fit the data. From a best-fit formula to predicting data—statistics at its finest. Armed with this information, Kepler believed there should be a planet between Mars and Jupiter. He was wrong . . . but he was kind of right too.

One hundred and forty years after Kepler’s death, another astronomer used the additional century’s worth of data and developments to create his own predictive formula, dubbed the Titius-Bode law. It, too, predicted a planet between Mars and Jupiter, as well as predicting the distance of a planet beyond Saturn.

Lo and behold, in 1781 astronomers found a planet beyond Saturn: Uranus.

The Titius-Bode law gained credibility. Soon, astronomers were searching for the “lost” planet between Mars and Jupiter. After conversing with the discoverer of Uranus, a Hungarian astronomer assembled a crack team of twenty-four fellow astronomers. Their purpose: to coordinate their efforts to systematically search for it. He dubbed them the Celestial Police.

One of these astronomers was an Italian priest in Palermo named Giuseppe Piazzi. As fate would have it, he had already discovered the lost planet before his invitation to join the Celestial Police arrived in the mail. As irony would have it, he discovered it by accident.

That New Year’s Day in 1801, Piazzi didn’t know what he was looking at. It looked like a comet, but it acted like a planet. He made his observations for a little over a month before he became ill. By the time he’d sent notes to other astronomers who could observe the unidentified orbiting object, it had been obscured by the sun. By the end of the year, Piazzi knew it would have almost completed its orbit and should be visible again . . . but he couldn’t find it. He’d only been able to make twenty-four observations—not enough to predict where it would reappear.

A twenty-four-year-old German child prodigy decided to tackle the problem of how to use scant data to create a formula to predict where this thing should be. In three months, Carl Friedrich Gauss developed “Gauss’s method.” (With just a bit of tweaking, Gauss’s method would later become the basis for global positioning systems, or GPS.) Using Gauss’s method, the Hungarian chief of the Celestial Police spotted the dwarf planet Ceres, the largest object in our solar system’s main asteroid belt between Mars and Jupiter, just a few days shy of the one-year anniversary of Piazzi’s first observation.

Six years later, Gauss abandoned pure mathematics (he didn’t think it was worth it, apparently) and became an astronomer himself. While pursuing astronomical calculations (literally), Gauss laid the foundation for the statistics we’re talking about here.

This was Gauss's way of quantifying the randomness of his and other astronomers’ observations. With the Plataeans’ ladders, being a foot off was good enough. When dealing with pinpricks in the night sky that constantly change position, he needed to be a lot more precise. Gauss needed to remove as many measurement errors as possible. His method was essentially a way to quantify what the Plataeans did instinctively. He called these variations the mean error; the statistician Karl Pearson would come to refer to this as standard deviation—a core principle Shewhart used to create Statistical Process Control.

Getting Quality Under Statistical Process Control

It’s eerie how much of Walter Shewhart’s life echoes Deming’s (or maybe the other way around). He was born nine years before Deming on a rural farm in western Illinois. In 1910, he enrolled at the University of Illinois at Urbana-Champaign. He earned his doctorate in physics there in 1917.

He taught physics in Wisconsin, but that changed when Congress voted to enter World War I after the sinking of the Lusitania. The US Army, Navy, and budding Air Force placed $22 million worth of orders—about $500 million today—for communications equipment with Western Electric. Spurred by a sense of patriotism, Shewhart left academia to join the company’s engineering department. He moved his family to Brooklyn and commuted to his office at Western Electric’s building at 463 West Street overlooking the Hudson. Of course, the vast majority of Western Electric’s manufacturing took place at Hawthorne Works. Sometime between the end of the World War I and 1922, Shewhart became interested (some might say obsessed) with improving production quality. He was convinced that he could take the mathematical statistics concepts he’d learned as a physicist and apply them to production processes. On a Friday in May 1924, he summarized some of his ideas in a one-page memo to his boss, George Edwards.

As Isaac Asimov is attributed as saying, “The most exciting phrase to hear in science, the one that heralds new discoveries, is not ‘Eureka!’ (I found it!) but ‘That’s funny . . . ’” That’s how I feel about Shewhart’s memo. It should have been a eureka moment. Instead, there was no fanfare. No ticker tape parade.

Nobody knew the history of manufacturing had made one of its most important turning points.

Shewhart created a statistical system to improve production quality. He took the idea of tolerance limits and flipped it on its head with statistics. Instead of looking at telephones as good enough or not—that black-and-white deterministic thinking the world was used to—Shewhart quantified the variation.

He could now tell the managers at Hawthorne what percentage of products fell statistically within their tolerance limits. Far more importantly, it gave managers a method to track variation. If you can track variation, then you can trace variation to better understand why a production line creates defects and detects them much earlier in the process.

In essence, Shewhart created a way to continually improve any manufacturing line. This was the first time factory managers had been given a tool to let them manage the uncertainty in production.

His method, Statistical Process Control, let managers compare variation across workers and machines. The more managers could find and fix the causes behind factory defects, the more products they could produce that fell within their tolerance limits, thus improving their overall quality. And higher quality meant less waste; less waste allowed manufacturers to do more and more with less and less.

Think of an automaker doing a recall because of a braking issue. Tens of thousands of cars all have to be brought into their local dealerships to get their brakes fixed, costing the car company a fortune. In general, it’s cheaper to produce higher-quality brakes in the first place than to fix a mistake after the fact.

Shewhart’s genius was taking statistics from academia—physics, astronomy, biology, etc.—and applying it to Hawthorne’s assembly lines. He wanted to cut down on how much Western Electric wasted. For the first time in history, we had finally stepped beyond a simple go/no-go approach to making things.

Tracking cold, hard numbers allowed managers to track patterns in the data. Shewhart’s method was a paradigm shift. It was the first example of anti-Taylorism, where using math and statistics enabled management to see defects as results of the process instead of the workers. Before this, most managers viewed their employees like Stradivari did his apprentices. If the product was bad, it must be the workers’ fault. Stradivari never stopped to consider whether the woodcutter had sold him wood from a diseased tree. Never stopped to see if the apprentices’ tools were sufficient. Never once considered that he himself might be a poor teacher. His reaction was to blame the worker.

By 1929, Shewhart had formalized this new method of tracking and tracing variation. Basically, he applied the scientific method to manufacturing. Before this, manufacturing was a linear process. You figured out what you wanted and how many. You made them. Then you inspected and threw away the defects.

Shewhart turned this into a cycle, what Ed would later call the Shewhart Cycle: Figure out what you want, make it, inspect it, figure out what caused the defects, go fix it, and then go through the whole cycle again, using feedback from your mistakes to continuously improve production quality.

Even after Deming tweaked the Shewhart cycle, he still referred to it throughout his life as the Shewhart wheel. Despite this, his students in Japan called it the Deming cycle. Today, you might recognize it as the “plan, do, check, act” method, or simply the PDCA cycle.

In his later years, however, Deming came to rename check as study. To his way of thinking, check was too much like the go/no-go inspection process of checking manufactured products. He believed the better term was study, which implied approaching the results with a scientific curiosity to investigate and understand why things turned out the way they did.

Why Does It Vary?