Root Cause Analysis

Root Cause Analysis Matthew A. Barsalou A Step-By-Step Guide to Using the Right Tool at the Right Time

Root Cause Analysis: The Core of Problem Solving and Corrective Action

Corrective action is the overall process involved with taking an identified problem and seeing that appropriate action is taken to resolve it. Within the corrective action process is a problem-solving process that finds and corrects the cause(s). The problem-solving process includes both a diagnostic phase and a solution phase, and it is the former that involves root cause analysis. Root-cause analysis is a process of drilling down to find causes of the problem so corrective action can be taken. By definition, corrective action means addressing causes rather than symptoms, but corrective action could be taken at the physical level only or also at the system level, depending on the criticality of the problem as well as frequency, cost, and risk.

(Similar to 5-Why)

Root cause analysis

Key Points

RCA can be performed to solve a problem or to identify improvement opportunities.
Teams should be used, ideally consisting of representatives from different departments.
An event or failure could have both a proximate and an ultimate root cause.
Flexibility is needed in the selection of tools.

Hypothesis testing

Key Points

A hypothesis should not make too many assumptions and should be simple and general. It must make a testable prediction to be of any use.
Occam's razor is a rule of thumb that states the hypothesis with the fewest assumptions should be selected when confronted with two competing hypothesis.
It is not possible to prove a hypothesis, only to support a hypothesis or reject a hypothesis. A well- corroborated hypothesis is one that has survived many attempts at refuting it.
Deduction goes from the general to the specific; deductive reasoning uses what is known to form a hypothesis. It uses previous hypotheses and models to form a new hypothesis.
Induction goes from the specific to the general; inductive reasoning uses observation to form a hypothesis. It uses facts and the observation of phenomena to form a hypothesis.

Procedure

Observe or collect the relevant data.
Analyze the data and look for relationships or patterns in the data.
Use the data to form a general hypothesis that goes beyond the data to make a prediction.
Test the hypothesis.
Use the test results as a basis for a new hypothesis if they fail to support the hypothesis. Perform confirmation testing if the results support the hypothesis.

Scientific method

Key Points

The scientific method is empirical/ evidence based.
A hypothesis is a key part of using the scientific method.
Quick iterations of the iterative inductive- deductive process can lead to the root cause.
The objective should not be to immediately identify a root cause but rather to move quickly through the cycles of PDCA to reject unsupported hypotheses and arrive at the root cause.
Avoid holding on to a pet hypothesis once it is unequivocally disproved. Evaluating an incorrect hypothesis is not bad; defending one is.

Procedure

Plan: Use deductive reasoning to form a tentative hypothesis based on observation.
Do: Assume the hypothesis is true for the sake of testing and evaluate it empirically, through either experimentation or observation.
Check: Compare the actual results against predicted results.
Act: Perform confirmation testing if the hypothesis was supported or move to the next cycle of PDCA and use induction to form a new hypothesis if the tested hypothesis was not supported.
Plan: Use the knowledge gained from the previous PDCA cycle to form a new hypothesis using induction.
Do: Assume the new hypothesis is true for the sake of testing and evaluate it empirically, through either experimentation or observation.
Check: Compare the actual results against predicted results.
Act: Perform confirmation testing if the new hypothesis was supported or move to the next cycle of PDCA and use deduction to form a hypothesis if the tested hypothesis was not supported.
Repeat the iterations of the PDCA cycle until a hypothesis is sup-ported, at which time confirmation testing should be performed.

Experiment

Key Points

A treatment, also known as an experimental run, is the set of conditions during an experiment.
A factor is a condition that affects an output, for example, temperature, material type, mixture, settings on a machine, or pressure.
The treatment variable, also known as an independent variable, is a factor that is manipulated by the experimenter to determine its effect or lack of effect on the response variable.
A response variable, also known as the dependent variable, is the result of the manipulation of the treatment variable.
The confounding variable, also known as the confounding factor, is a source of noise.
Noise in an experiment is an uncontrolled and potentially unknown factor that influences the experimental results.
Precision is the closeness of measurements to each other.
Accuracy is the closeness of a measurement to the true value.
Blocking reduces variability and increases precision by spreading confounding variables across the experimental results. (An uncontrolled variable is distributed across all treatments by ensuring all experimental groups or blocks contain the confounding variable, thereby decreasing variability and increasing the precision of the experimental results.)
Randomization increases the accuracy of experimental results by canceling out the influence of noise.
Replication is the repetition of an experiment to increase the accuracy and precision of the results.
Operational definitions are clear quantitative descriptions of terms using tests or measurements to define the terms.
Failing to check the baseline may result in attributing changes in the response variable to the setting of the treatment variable when no actual relationship exists and the response variable would have changed regardless of the setting of the treatment variable.
Blind tests may be needed to increase objectivity.

Procedure

Create a test plan based on a hypothesis. The predicted result of the hypothesis is the response variable.
Determine the treatment conditions by establishing the treatment variable or variables.
Identify potential confounding variables and establish a method to eliminate, control, or minimize them. Blocking and randomization may be useful here.
Ensure all terms are written as operational definitions.
Establish the baseline if necessary or possible.
Perform the experiment.
Replicate the experiment and compare the results; a large difference is an indicator that variation is present and more replicates are needed.

8D report

Key Points

The 8D report provides a brief, but detailed, report on root cause analysis activities as well as the root cause.
It can be used for internal or external failures.
An 8D report is typically used for the investigations of failures, but the same format can be applied to improvement activities.
Different companies may use different names for the eight steps; however, the steps themselves are the same.
The fields on top of an 8D report should be customized to fit the needs of the company using the 8D report.

Procedure

Start the 8D report by filling in the fields at the top of the document and forming an interdisciplinary team.
Describe the problem.
Decide if containment actions are necessary. If so, assign somebody to implement them and report on the results of the containment actions.
Identify the root cause of the failure by investigating the part or process that failed. The investigation should be supported by the use of quality tools.
After the identification of the root cause, the corrective actions should be described. Trials should be performed to ensure that these actions will be effective.
The implemented corrective actions need to be described.
Actions must be taken to prevent a reoccurrence of the issue.
The team needs to be congratulated, and the report should be closed.