This Billionaire Went Data-Driven Before It Was Cool - Data and Decision #8
This Billionaire Went Data-Driven Before It Was Cool - Data and Decision #8
We talk how Charles Koch being a successful data nerd before most of us can spell "machine learning".
Hello there!
Last week, we know that one way to build data-driven organization is by having strong CEO who also believe in data at the top. This leads us to one of the insights that I want to share with you on this issue: how billionaire CEO Charles Koch instill metrics-driven mindset in Koch Industries. Other thing that I want to share with you is a good tweet about Value-at-Risk (VaR), or what Forbes called, “Formula from Hell”.
Oh, I've also changed the name to “Data and Decision” to better reflect the new focus I have. So, there's that.
Alright then, shall we get down to business?
2 Insights
Koch’s Way to Become Data-Driven Company
To say that Koch Industries is a unique company would be an understatement.
Under the leadership of Charles Koch, this company was among the first to purchase and utilize computers to support their operations. Let me remind all of you that during that era, computers were massive and often occupied entire rooms. And their primary users were government agencies like NASA.
Koch Industries also invited a young business consultant named Michael Porter, who had just published his legendary book “Competitive Strategy,” to give a TED Talk of sorts to their employees.
However, Koch Industries also has darker chapters in its history. In its very first year, the company faced a battle with its labor union over one of its most critical refineries. They faced fraud allegations initiated by the US Senate and investigated by the FBI. And we won't even have time to delve into the political track records of both Charles Koch and David Koch.
But let's set all of that aside for now. Whatever you think about Koch Industries, what we want to know right now is how data can help small company that started in oil refinery business to become one of biggest conglomerate in modern US.
Operational and Strategical Knowledge
He built a company that learned constantly from the world around it and prized information discovery above almost everything.
(…) In the face of unprecedented market volatility, Charles Koch and his team adopted a strategy that would inform Koch Industries for decades. It relied on deep analysis and information gathering.
Kochland written by Christopher Leonard
Now, before I delve into this further, let me address a hypothesis that I had challenged after reading the book “Kochland”. It was my belief that data could only support what I referred to as “operational” day-to-day activities and had limited, if any, role in improving a company's overall strategy.
This bias could have stemmed from various sources that I hadn't even realized I was influenced by. At one point, I read in Hamilton's “7 Power” that data alone isn't a moat (or, to be more precise, a 'recommendation engine' built from data isn’t a moat), suggesting that companies could not rely on accumulating more data to enhance their “Power”. Another factor contributing to this bias might have been my previous notion that data could only come into play to support execution, in complete contrast to what people talk about 'Strategy' with a capital 'S.'
However, Koch Industries showed me a different way to treat these data.
I already knew that Koch Industries had implemented ideas from Deming, one of the famous management gurus from the USA. But the article I read didn't just mention that Koch used Deming to become data-driven; it revealed an entirely different philosophy. Koch had established what they referred to as a 'development group' with a focus on acquiring companies whenever and wherever Koch's employees saw potential. This 'development group' became critical in how Koch conducted its business and formulated its overarching strategy.
Hence, I believe it's essential, before we proceed any further, that I explain what I mean by 'operational' and 'strategical' knowledge, the term that you might already see in the subtitle.
(((Yeah-yeah, I know. Spending your afternoon by reading definition in random newsletter isn’t fun stuff. But bear me with here)))
To be more precise, let's define 'operational' knowledge as 'knowledge that can be executed and its outcomes will be visible within a few minutes, hours or days at most.' For a clearer illustration, think of operational knowledge as something akin to driving instructions: 'If you press the brake pedal, the car will stop.' You know what to do, and the feedback loop is rapid (the car stops). There's a straightforward connection between 'brake' and 'car stop.'
In contrast, 'strategical' knowledge, as I define it, is 'knowledge that, after executing it, still involves a degree of uncertainty regarding whether the anticipated outcome will be achieved.' Expanding on the previous driving example, if I were to advise you, 'You should avoid St. James Road on Monday mornings and use Oak Road instead to avoid being late,' you'd understand that this type of knowledge doesn't guarantee an exact outcome. You might still arrive late, even if you follow my advice and take Oak Road.
In essence, I referred to 'operational' knowledge when dealing with domains where causation is strong, while 'strategical' knowledge tends to be more uncertain regarding its impact.
These two forms of knowledge, at least in my perspective, feel rather distinct. Initially, I believed they couldn't be combined, as they appeared to have little or no overlap.
Now, you might wonder why I've gone to such lengths to explain this concept. That's because this is the recurring theme that I observed throughout the book (at least so far).
In a concise statement: Koch used the information they gathered to construct 'operational' knowledge but then leveraged it to compile 'strategical' knowledge and acquire companies they deemed suitable.
Dubose, Deming, and Koch’s Way to Cut Cost
Our main character for this section is Phil Dubose, an employee at Koch Industries. In this particular story, Dubose had just been promoted in 1982 to oversee Koch's marine operations in the Gulf of Mexico. He was now responsible for a fleet of barges that traveled between terminals, collecting crude oil and shipping it to refineries in Texas.
However, despite his new position, we learn from the book that Dubose was deeply worried about the promotion. There had been two previous managers of the marine unit, and both of them had failed to turn a profit.
“If it failed again, I was going to go down with it,” Dubose recalled. But, Dubose was determined to make the shipping barges turn a profit. He knew he had one tool to help him do this: the charts of W. Edwards Deming.
What Dubose did for his division essentially involved two key actions:
Dubose used charts to visually represent the current performance of each ship, including their overall trends and variations.
Dubose actively shared and distribute this knowledge and insights with each of his skippers (the heads of the ships) and held them accountable for making improvements.
I'll provide a direct excerpt from the book regarding the first action:
The fleet Dubose oversaw initially consisted of five large barges. They each carried about 8,500 barrels of oil. Each barge had a skipper and crew who lived on the craft while it traveled from port to port.
The first matter of business that Dubose focused on was keeping costs down. Fuel was the largest cost the barges incurred.
(…) The tools from Deming helped Dubose go even further. Of all the charts he learned to make, he found that by far the most useful was called a run chart. Even decades later, he’d talk about run charts as if he were discussing a cherished family pet. “The best chart out of all of them . . . is that old-fashioned run chart. It’ll tell you where you’ve been and where you’re going,” he said.
A run chart broke down all the costs that a barge would incur. It had a separate category for each cost: groceries, fuel, maintenance, ship damage, and supplies. The run chart allowed you to track these costs as they shifted from month to month, letting you see “where you’ve been and where you’re going.” Dubose was taught to look for cost spikes.
The reason was simple: you figured out what caused costs to spike, and you avoided it. Then you figured out what caused costs to fall, and you replicated it. (emphasis added)
This was follow up by the “share and distribute knowledge” part that I mentioned above.
The critical part came next. Dubose printed run charts for each vessel and posted them in the skippers’ cabins. Each skipper could then see for themselves where they were running up costs and where they were saving money.
Dubose turned each skipper into his own manager. Skippers were free to make their own decisions based on the run chart.
Then Dubose went further. He started tracking the profits and losses for each barge. This made each skipper a small-business owner and each barge a small business. The skipper had all the information he needed to boost profits and the freedom to act on that information. And Dubose had total visibility into his fleet; he knew which ships were losing money and which were making it.
“It got to the point where the boats were competing against each other. I was just sitting back like a big old Cheshire cat in a tree,” Dubose said. Using data to drive changes at the level of each barge, Dubose boosted profits in the marine unit overall. His profit margin reached 33 percent.
(..) As he boosted profits, Dubose was given more freedom and more resources. He added more ships, buying larger barges that could ship forty thousand barrels of oil at a time.
What Dubose do, I believe, is an example of “operational knowledge”. To increase margin, Dubose know that he need to reduce the cost. But before that, he must understood the current performance and its trend (run chart). Knowing that “operational” side might be suitable to be decentralized, he share the knowledge about this to his team but use it also as a measurement of their performance.
Classic textbook of “optimized by data” in my opinion.
Koch’s Development Group: An Acquisition Machine
Koch has always had an appetite for expansion; it's an integral part of their identity. Sterling Varner, one of the directors and Charles' right-hand man, mentions their "instinct to scan the market for new opportunities." This drive is also partly ideological, rooted in Koch's belief in the dictum of Austrian economists Hayek and von Mises, which states that "markets never stand still." They believe the status quo never survives, as markets continually evolve and transform.
That’s the “why” for explaining Koch’s appetite. Now, the “how”.
The book provides an example of how Koch leveraged data that they compiled internally. They used it to assess potential companies and take action. Koch's Development Group even competed with private equity firms from Wall Street.
From the book:
Koch’s development group would become one of the largest, most effective deal-making machines in the United States.
(…) Charles Koch quietly built a private equity firm inside his offices in Wichita that would rival anything created on Wall Street. In the earliest days, in the 1980s, virtually nobody outside Koch Industries headquarters knew that the development group existed.
The development group made its first major deal in 1981. It came along thanks to Bernard Paulson, the head of oil refining. He had spotted an opportunity in part because of the computer models that he had perfected to help him run the Pine Bend refinery. The data helped Paulson determine exactly which units he should run, what products he should produce, and which markets he should sell into. The computer models gave Paulson a kind of X-ray vision into oil markets. Now, Paulson turned that vision outward, toward his competitors.
Koch Industries sold a lot of crude oil to a refinery owned by Sun Oil in Corpus Christi. But Koch didn’t just collect money when it sold crude to Sun Oil. It also collected intelligence.
Bernard Paulson’s team knew how much oil Sun was purchasing, and what kind of oil. Then he learned who Sun’s customers were, and what those customers paid for Sun’s product. Paulson began using his computer models to study the market that surrounded Sun Oil’s refinery. He studied what equipment was inside the refinery and at what volume that equipment could process oil. He learned what products Sun was making and at what volumes. He learned where Sun was selling its products and at what price.
(…) But Paulson saw something in Corpus Christi that even the refinery’s owner did not seem to appreciate: he saw that a market opportunity was being wasted. The Sun Oil refinery had equipment that could process oil and turn it into a petrochemical called paraxylene (pronounced pair-uh-ZIE-lene). Paraxylene was one of those products that Koch Industries excelled at making and selling: it was obscure, difficult to produce, and used in one form or another by virtually everyone.
Paraxylene was the raw material for synthetic fibers and materials like dimethyl-terephthalate acid and purified terephthalic acid. These chemicals, in turn, were used to make things like polyethylene terephthalate and saturated polyester polymers. Most people have never heard of these chemicals, but they are the building blocks for plastic containers, bottles, drapes, upholstery, clothing like polyester suits, electrical insulation, and photographic film. Paraxylene was something that everybody bought without knowing it.
And demand for paraxylene was growing.
(…) If Koch bought the Corpus Christi plant, Paulson realized, the acquisition would open up an entirely new market for the company: the market for paraxylene and other petrochemicals.
(…) In September of 1981 Koch Industries paid $265 million in cash for the refinery, and Paulson immediately started expanding it. He more than doubled its paraxylene output. He bought a used hydrocracking tower from a refinery in Europe and had it shipped to Texas, bragging to Charles Koch that he bought the tower for 40 percent of what it would cost “off the shelf.” Koch Industries became one of the largest paraxylene producers in the United States.
So, What’s the Practical Things After Knowing This?
I hate to leave you with a 10-minute read and having none practical aspect that the reader can grab and apply it. However, to be intellectually honest with you, I feel it's still too early to distill practical insights from these observations. Currently, only the "operational knowledge" aspect offers something tangible for our daily work. I also have the ethical concern that I too focus on “interestingness” and can’t distill it to “useful method”.
Nevertheless, two observations have been stuck in my mind so far:
If you can understand what a company does with its data and if it match with above pattern of "expand using operational knowledge", you can confidently know that the CEO have strong appetite of data. Also ideally, you should work at this type of company.
One thing that I had in my gut is we still overlook the important of “instinct to scan the market for new opportunities” that Koch’s had. I think this is another way to confirm my hypotheses that before doing all data collection, reporting, and analysis, the company must believe that data is valuable. It’s akin to religion compared to scientific pursuit. And I don’t know why.
Is there anything else that has stuck in your mind after reading my thoughts above?
“Formula from Hell” and Why Nassim Taleb Want to Ban It
Part of my reason for changing the name of this newsletter to include "Decision" is quite philosophical: insight is useless if we don’t act on it. We only know if it useful by comparing our expectation with the actual outcome. After all, the only useful pursuit is to add more useful knowledge that already pass the “authentication method”, usually market or reality.
This will provide a nice segue way for our next insight: never in the history of human existence has a single formula (or, using my previous term, "knowledge") caused so much pain in the world.
Valeriy, one of my favorite stats tweeters out there, recently reshared a post/tweet about Value-at-Risk (VAR), a formula that was quite famous until the 2008 financial crisis. I remember one day pondering just how bad VAR really is, and I came across a video of Nassim Taleb mentioning that he wanted to ban VAR (along with other conventional measurements used in the banking industry). My younger self (I remember) chuckled a little bit watching someone as smart as Nassim proposed something so specific. On second thought, I realized that illustrated how deeply Nassim understood the root cause of this crises.
This tweet contains numerous banger quotes (and you can read some of the articles too), but here's my personal favorite:
The “masters of universe” relying on XIXst century Gaussian distribution trying to describe how markets work using normality assumption - what could possibly go wrong.
That’s all! Thank you folks.