The Wager
For years, I have occasionally put a hundred-dollar bill on a table beside an unopened package of Starburst.
I tell the people around me that I can break the package roughly in half and predict the colors of the two candies that will appear at the break. They can put up ten dollars against my hundred. Or five. Sometimes I have offered to let them risk a single dollar.
People still hesitate.
The demonstration began with an observation I made years ago while running errands with my two sons. I noticed that Starburst packages seemed to have a recurring pattern. I started paying attention, then predicting what I would find.
I have seen the pattern change, so I would never claim that Starburst guarantees a particular arrangement or that I know how the candy is packaged.
If you assume four possible colors, independently and randomly distributed, predicting two specific positions appears to be a one-in-sixteen proposition. My confidence in the wager comes from evidence that the assumption is wrong.
I cannot see the process, but I can see evidence of it in the product.
A repeatable process can also consistently produce the wrong thing. Repeatability matters only when the result being repeated matters.
I learned that distinction after I became engineering manager at Powder River and took responsibility for research and development.
Steel was expensive, our cattle-handling equipment was heavy, and I believed we could use less material without compromising performance. The R&D manager and I argued about steel gauge regularly.
I went back to my engineering textbooks, performed the analysis, and used finite element analysis to examine the structure. Everything I saw reinforced my belief that we were overbuilding the product.
He disagreed.
Eventually, instead of continuing the argument, he invited me to work cattle with him at his ranch.
We took a new tub and alley and set it up. It went together beautifully. It was also heavy, and while we assembled it I continued thinking about all the steel I believed we did not need.
Most of the cattle moved through without incident.
In the last group was a large, fully grown bull that did not want to move from the tub into the alley toward the chute. I was at the edge of the tub and slapped him on the hind end to encourage him forward.
The bull reared and kicked the sheet directly in front of me.
There was an enormous bang. The sheet shook hard enough that I fell to the ground. For a moment, I was convinced I had been kicked and seriously injured.
Then I realized I was fine.
The sheet had a small bend in it.
I went to the R&D manager and apologized.
“I think you have exactly the right gauge material in that sheet. Let's not change a single thing.”
The equations had not stopped working. I had been optimizing against an incomplete understanding of the requirement.
At Powder River we sometimes referred to institutional knowledge as sacred cows. We also liked to say that sacred cows make the best steaks. “We've always done it this way” was not a sufficient reason to keep doing something.
A future engineer might find a better geometry, material, or manufacturing method that provides the same protection with less weight and cost. The gauge itself was not sacred. What mattered was understanding what the equipment had to withstand before deciding how much strength was enough.
After that day, I understood the basis for the R&D manager's judgment better. I also became much more interested in testing products where they would actually be used. I attended ranch tests regularly and encouraged us to find cattle that were less docile rather than arranging demonstrations where everything behaved perfectly.
Eventually, I began teaching an hour-long class to our dealers about the engineering and construction behind the equipment. Then we took the class to a ranch.
Dealers worked the tubs and chutes themselves. They felt what happened when cattle loaded the equipment unpredictably.
I taught the program to more than nine hundred people. I sometimes included managers and long-term employees from our own organization as well.
Organizations accumulate knowledge. Some of it ends up in drawings and specifications. Some becomes work instructions, fixtures, tolerances or tests. People do not have to rediscover every lesson every time they perform the work.
But a drawing can preserve the answer without preserving the argument that produced it. A procedure can tell someone what to do without telling them what failure it is intended to prevent.
Eventually someone will challenge the answer.
Old processes become obsolete. Specifications can be wrong. Protecting everything inherited from the past is not wisdom.
Neither is removing something merely because its purpose is no longer obvious.
The challenge is to preserve enough understanding that people can tell the difference.
When the next person comes along with a good idea for making something lighter, faster, simpler or cheaper, I hope they challenge what came before.
Just make sure they understand the wager before they place the bet.
← Essays