AAOM Handbook

INTERNAL

AI.13 Validate or Reject Potential Causes Context

Brainstorming should result in the identification of many potential causes for an issue. Even after these have been organised and rationalised there may still be a large number of potential causes. There is a natural tendency at this point to want to rush into action, that is, make some change to the process based on the causes we have listed. However, at this time none of these potential causes has been proven. A change made where there was no issue will generally make the process perform worse, not better. Before any control action is taken we need to validate that a potential cause is both a real and significant contributor to the process issue we are investigating. We ranked the potential causes to help reduce the workload required to validate them. We start the validation from the highest ranked cause and proceed until we have confidently identified the significant contributing/root cause(s). Some common techniques for validating the contribution of a probable cause to an issue include;  Determining the correlation between the KPI in which the issue was observed and a control measure believed to influence the KPI.  Producing a Pareto diagram of the measured (actual) impact of the common causes affecting a KPI or control measure.  Designing and conducting an experiment to test an hypothesis on the cause of the issue. This task is to specify and collect the data necessary to validate or reject the most probable causes. Purpose To validate the contributing/root cause(s) for a process issue. Quantity One specification for validation data for each probable cause. One set of validation data for each probable cause. Quality Consider the following;  Type of data to be collected (is it a measured value, a characteristic, a cause of variation),  The decision that is required from the data (is the cause real, what is the extent of influence of a cause on a variable),  Characteristics of the data,  Potential sources of the data,  For measured values the quantisation level, based on the minimum value of change that is significant for the variable to be measured.

© McAlear Management Consultants 2007

Page 25 of 45