Publications

This first volume teaches the fundamentals of flow problem-solving and the basic principles of Lean Management. The book is intended for students and instructors of Lean Six Sigma, as well as managers who wish to move beyond simply acquiring formulas and develop their practices on a solid theoretical foundation. While respecting the historical roots, spirit, and established terminology of Lean Six Sigma, the book strives to identify, unify, and formalize its underlying fundamental principles.

This second volume details the implementation of the approach within companies and the generic DMAIC methodological protocol. Based on an original definition of statistical problems, it teaches the fundamentals of solving "discernible" statistical problems and the basic principles of Six Sigma. The book is intended for students and teachers of Lean Six Sigma, as well as operational improvement practitioners who wish to move beyond simply acquiring formulas and develop their practices based on a solid theoretical framework.

This third volume teaches how to analyse and reduce problems whose causes remain indistinguishable in individual cases. Drawing on an original analogy with optics, it explains how to calibrate analytical resources to the improvement sought. It shows how to use hypothesis testing to identify useful improvement levers, assess realistically achievable gains, and determine whether proposed solutions can meet the target. The book is intended for Six Sigma practitioners and users of statistical analysis who wish to fully exploit the decision-making potential of small datasets.

Abstract :

%GRR and Number of Distinct Categories (NDC) are widely used as measurement system capability indicators. However, the usually accepted calculation formula of the NDC is assigned a coefficient whose origin have led to the coexistence of two different and misleading interpretations of the NDC. We recommend to remove this coefficient in order to stick to the measurement system resolution definition or equivalent signal-to-noise concept. We also propose a realignment of the %GRR and NDC acceptance thresholds.

Abstract :

As part of the Measure phase of a Six Sigma (DMAIC) project, it is key to determine the sample size appropriate given the statistical precision needed on the estimation of the process capability (defined here as the expected proportion of defective units). Six Sigma practitioners are generally taught to use either some rule of thumb (skipping so the precision requirement) or a sample size formula allowing to specify the required precision. While such a formula is provided for binary data, when dealing with defect counting data, the usual formula taught is about estimating the mean number of defects per unit and not the proportion of defective units (i.e., the defective rate). This paper elaborates a formula and derived tables for calculating the sample size required to achieve the needed statistical precision on the defective rate when working with a defect counting metric. Reversely, the precision obtained for a given sample size is calculated. The results are then compared to those for binary data and it is suggested that the sample size calculation for binary data is worth to be considered as an acceptable easier-to-use alternative. The challenge to obtain decent precisions due to the sample size requirements when operating at low defective rates is also highlighted.

Abstract :

As part of the Measure phase of a Six Sigma (DMAIC) project, it is key to determine the sample size appropriate given the statistical precision needed on the estimation of the process capability (defined here as the expected proportion of defective units). Six Sigma practitioners are generally taught to use either some rule of thumb (skipping so the precision requirement) or a sample size formula allowing to specify the required precision. While such a formula is provided for binary data, when dealing with continuous data, the usual formula taught is about estimating the mean and not the defective rate. Other alternative approaches are about calculating the sample size based on precision on process capability indexes designed for manufacturing industry. However, these technical indexes are often not relevant for most users and managers not working in specific production environments. This paper elaborates solutions for calculating the sample size required to achieve the needed statistical precision on the defective rate when working with a continuous metric in both cases of one and two specification limits. Reversely, the precision obtained for a given sample size is calculated. The challenge to assess high Sigma performance levels because of the required sample size is also highlighted.