Statistics Based Kaizen

Statistics-based Kaizen, SBK.

SBK is a term used in Quantum TC Consulting to refer to improvement needs generated by awareness of predictive potential gap and its derived success. Statistics-based kaizen in Quantum TC is a combination of three tools: 2 pertaining to the Six Sigma world and one coming from Lean Management culture.

In Six Sigma there is an analysis to determine the level of agreement existing in a measurement system by attribute and it is known as Kappa. The Cohen’s Kappa index establishes the consistency of an appraiser about and acceptance criterion, the agreement between appraisers for the same criteria, and the concordance between all appraisers and the standards previously established.

Also, in Six Sigma statistics there is a calculation of probabilities, given a set of data. It can be calculated the probability of occurrence of a dependent event given the prior occurrence of another event, the independent event. This is called the Bayesian probability or simply conditional probability: the probability of an event B given the prior occurrence of event A.

On the other side, in Lean Management, it is used a skill matrix to determine both process flexibility in terms of amount of operators accumulated knowledges, and operator flexibility in terms of amount of process accumulated knowledges. Skills matrix differs from cross training matrix by establishing a level of acquired skills in each operation, by each operator.

Given the fact that Cohen’s Kappa study presents the details of matched inspection between appraisers and the fact that 2 or more appraisers are used for the study, it is possible to calculate the probability of appraiser B finding a defect that appraiser A had let go. This conditional probability is especially useful when appraisers are taken from different levels of inspection in the value stream. For instance, appraiser A could be inspection realized directly on production, while appraiser B could be final audit inspection before delivering to customer. When used for this purpose it is often called PPOS: Predictive Probability Of Success.

Predictive probability of success (PPOS) is a statistics concept particular used in the pharmaceutical industry, but applicable in all other manufacturing sectors, and to a great extent in services as well, to support decision making toward improvement needed or not needed. PPOS is the probability of observing a success in the future based on existing data. PPOS is the fundamental performance indicator for Statistics-based kaizen. It is one type of a Bayesian means, by which the management can determined the data's likelihood over possible future responses when skills matrix oriented to defects are used.

This means that once the probability is calculated, the next step is to adapt the skills matrix concepts to the process, changing operations by defect types, and training appraisers to acquire different skills levels (untrained, learner, practitioner, developer, coach) in a specific defect identification.

Let us suppose that a Kappa between John and George is found to be 0.6532, which in terms of measurement system is not acceptable, and the probability of George, being the final quality control auditor, finds a defect given that John, the production line inspector, did not find, is 70%. The predictive probability of success in this case is only 70%, leaving a statistics-based kaizen up to 30%. These 30% must be compensated by developing both George and John defects recognition and segregation skills through special training using the skills matrix to calculate their actual and predictive flexibility in defects identification.