TRY THIS TODAY
Keep It Simple
The case for adding Occam’s razor to your root cause analysis toolkit
by Matthew Barsalou
When performing root cause analysis, it’s helpful to form a tentative hypothesis. Your hypothesis might be wrong, but it gives you a starting place—something to evaluate. If new information is gained while evaluating the hypothesis, it can be rejected if it’s found to be incorrect.
Suppose, for example, you discover that some machined parts are rusted. You might hypothesize: “The parts were wet with coolant when they were sent to the warehouse.” This can be evaluated to determine whether your hypothesis is true. If the failure is sporadic, a simple experiment can be performed by intentionally wetting sample parts with coolant, storing them in the warehouse and periodically checking them for rust.
A hypothesis has a better chance of being correct if it makes fewer assumptions.1 This hypothesis, for example, is too complex: “The parts rusted in the warehouse because they were wet with coolant when they were sent to the warehouse. They were wet because the machine that blows off the coolant wasn’t set properly. The machine wasn’t set properly because the work instruction wasn’t clear.”
This may be the chain of causality, but starting from the final effect and reasoning backward results in too many assumptions. Each item could be its own hypothesis, but together, they make a weak hypothesis. Only one wrong assumption will break the chain of causality.
The problems with a complex hypothesis are well-illustrated by the Linda problem: “Linda is 31 years old, single, outspoken and bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations.”2 The authors of the problem ask whether it’s more likely that Linda is a bank teller, or a bank teller and a feminist.
It’s easy to imagine Linda is a feminist, so many people conclude she’s a bank teller and a feminist. The odds of Linda being a bank teller, however, are greater than the odds of her being a bank teller and a feminist.
Think of it this way: The odds of Linda being a bank teller are 40%, and the odds of her being a feminist are 70%. The odds of Linda being a bank teller and a feminist are only 28%, while the odds of her being a bank teller are still 40%. The more complex hypothesis, “Linda is a bank teller and a feminist,” requires more assumptions that must be true for the hypothesis to be correct.
There is a simple solution to the problem of overly complex hypotheses: Occam’s razor. According to this problem-solving principle, if two hypotheses explain the data equally well, the simpler hypothesis is usually better.3
As illustrated by the Linda problem, a hypothesis that makes fewer assumptions is more likely to be correct than one that makes many assumptions. This doesn’t mean additional assumptions can’t be evaluated as different hypotheses—one of many hypotheses could and should be correct. But combining the assumptions into one convoluted hypothesis increases the chances of the hypothesis being wrong.
- Willard V. Quine and Joseph S. Ullian, The Web of Belief, 10th edition, Random House, 1978.
- Amos Tversky and Daniel Kahneman, “Judgments of and by Representativeness,” Judgment Under Uncertainty: Heuristics and Biases, Cambridge University Press, 1982, pp. 84-100.
- Carl Sagan, The Demon Haunted World: Science as a Candle in the Dark, Ballantine Books, 1996.
Matthew Barsalou is a certified Six Sigma Master Black Belt located in Germany. He has a master’s degree in business administration and engineering from Wilhelm Büchner Hochschule in Darmstadt, Germany, and a master of liberal studies from Fort Hays State University in Hays, KS. Barsalou is an ASQ senior member and holds several certifications.