Nunnally and McConnell's objective is to fill a void ‘on the basic building blocks of understanding and reducing variability generally, and Six Sigma in particular, that was specific to the pharmaceutical industry’. The focus is on only part of the industry – the process of manufacturing pharmaceutical products, and related analytical measurement processes. Even for manufacturing, other processes such as the financial and procurement processes are not considered. There is no attempt to show relevance to Research and Development processes, other than a passing reference to Design for Six Sigma. Thus, the authors reinforce a common misapprehension that Six Sigma is only for manufacturing, whereas adopting a Six Sigma culture can bring benefits for any processes.

Given their objective is to understand variability, they rely heavily on statistical thinking – the Deming Red Bead and Funnel Experiments are discussed in detail and pictures of distributions and run charts with limits abound. These basic statistical concepts are not explained and there are many other unexplained terms, for example analytical measurement/validation constructs such as ‘intermediate precision’. For those with little statistical expertise wishing to learn about Six Sigma, much will be annoyingly abstract and academic. In contrast peripheral facts are given in detail, for example that the points of inflection in the Normal Curve are at one standard deviation σ from the mean. For some reason this fact is much loved by some Six Sigma trainers. Here and later when the formula is stated, there is little justification of σ as a useful measure of variability and confusion over whether to use n or (n-1). In addition, there are many editorial errors.

Process capability and the Taguchi loss function are mentioned, but only one statistical method is presented, the Statistical Process Control (SPC) or Shewhart Chart. Grant and Leavenworth ‘Statistical Quality Control’ (1980, McGraw-Hill, no edition given) is ‘their principal statistical reference’. Their approach to SPC charts follows that used in my copy of an earlier edition of this book [1]. They comment that other methods exist, but without advice on where these can be found. SPC charts are presented as the data analysis tool, not for process control. Different charts are assessed in an academic way with no perception that the practical situation defines the data available and thus the correct chart to use or that SPC is a two-stage process – first creating statistical control and then setting limits. So even for SPC charts, I feel that they do not achieve the aim of giving ‘basic building blocks of understanding and reducing variability’.

The other chapters cover standard topics, including the case for Six Sigma, history and development, and examples from pharmaceutical manufacturing. With the exception of regulatory compliance aspects, I found the last lacking detail and context, though I recognise difficulties in doing so while preserving confidentiality. These chapters, particularly the stress on the cultural and emotional aspects, could be useful in introducing Six Sigma to statisticians and statistically literate scientists in development who need to work with manufacturing. However, there are other books that I would rather recommend, including [2].

Though this is not simply yet another Six Sigma ‘cookbook’ of tools and techniques, I feel that they are far from achieving their aim given above or that this book ‘provides the basis of a complete operating policy’.

Finally, it is interesting that neither author is a statistician, nor apparently considered it wise to involve a statistician in writing a book that relies so much on statistics. I leave it to my fellow statisticians to ponder on this.


  1. Top of page
  • 1
    Grant EL, Leavenworth RS. Statistical Quality Control (4th edn). McGraw-Hill Kogakusha: Tokyo, 1972.
  • 2
    Gygi C, DeCarlo N, Williams B. Six Sigma for Dummies. Wiley: Indianapolis, 2005.