BACK TO BASICS
Shedding More Light
The scientific method can help answer difficult questions in root cause analysis
by Matthew Barsalou
The scientific method is used in science; however, it’s also appropriate when performing a root cause analysis. Using the scientific method can help ensure that the true cause of a failure is identified. Identifying the cause of a failure is essential because corrective actions will be ineffective if they are not based on the root cause of the problem.
According to Stephen Tramel, there is no one scientific method. There are, however, six steps that are commonly used to test and evaluate the hypotheses (see Table 1).1
The first and last steps are empirical: They are based on observations by using senses such as touch, hearing or sight, and perhaps through measuring devices such as calipers.
Observations are made and data collected during the first step. Next, data are analyzed. In this example, an analysis of the data indicates that all failures are occurring on parts produced on night shift. Then you must generalize beyond the available data to form a hypothesis. In this case, the hypothesis is: "Only parts produced during the night shift are failing."
The fourth step is assuming the hypothesis is true for the sake of testing. Next, test results are predicted based on the assumption that the hypothesis is true. The hypothesis is tested during this final step. Parts from the day and night shifts are completely inspected for a week. Only night-shift parts should have failures if the hypothesis is correct. In this example, there is an approximately equal number of failures between the two shifts, so the hypothesis is rejected.
George E.P. Box, Stuart Hunter and William G. Hunter explain that induction and deduction are used in forming and evaluating hypotheses.2 To form a hypothesis, you start with what is known, including what has been observed, and any known theories and uses. The hypothesis is tested. If it fails the testing, induction is used to go from the results to a theory that fits the results, and a new hypothesis is formed and tested.3
This process can be repeated until a hypothesis is robust enough to survive testing. Using the earlier example, you may hypothesize: "The failure occurs at an equal rate between shifts, but the night shift was better at detecting the failure."
The root cause still has not been found, but investigators now better understand the issue and are closer to finding the root cause because they are no longer examining nonexistent variation between shifts.
The scientific method can help answer difficult questions and also can lead to new discoveries. In addition to be being useful for researchers and scientists, the method can apply when investigating a quality failure.
1. Stephen Tramel, "Explanatory Hypotheses and the Scientific Method," Ways of Knowing in Comparative Perspective: The WKCP Companion and Anthology, Copley Custom Textbooks, 2006.
2. George E.P. Box, Stuart Hunter and William G. Hunter, Statistics for Experimenters: An Introduction to Design, Data Analysis and Model Building, second edition, John Wiley & Sons, 2005.
Matthew Barsalou is a statistical problem resolution Master Black Belt (MBB) at BorgWarner Turbo Systems Engineering GmbH in Kirchheimbolanden, Germany. He has a master’s degree in business administration and engineering from Wilhem Büchner Hoschschule in Darmstadt, Germany, and a master’s degree in liberal studies from Fort Hays State University in Hays, KS. Barsalou is an ASQ member and holds several certifications.