The Statistician in Quality Assurance

Quality of goods and services is important to all customers. Therefore the person who takes responsibility for assuring the quality is critical to both the organisation and its customers. In many professional firms it might be a senior partner while in many large manufacturing companies it might be a senior engineer. Often the role of that person is to ensure that standards and specifications are met.

It may be a surprise to some that statisticians are increasingly filling this role of providing quality assurance. The reason is simple, reducing variation in a product or service is often the most cost effective way of reliably meeting specifications. Understanding variation is the key to this and is also the statistician's chief skill. It is perhaps not surprising that many of the prominent names in the quality movement - Shewart and Deming to name just two - described themselves primarily as statisticians.

Many of the techniques in wide use in quality assurance are basically statistical but have been simplified for use by non-statisticians. These include such tools as control charts and histograms. However an experienced statistician has the greater depth of understanding to handle the unexpected and to take advantage of unusual situations.

An example of this is a recent project where, as statisticians, Data Analysis Australia provided certification for a reference material, or standard, to be used by mineral laboratories as a part of their own quality assurance procedures. These mineral standards are a critical product of our client, Western Mineral Standards, with laboratories around the world using them to calibrate their results, in this case for measuring gold concentrations in mineral samples. Several standards were being developed simultaneously and as this was a new subsidiary and area of business for Western Refractories, extra care had to be taken in the review of the processes used to ensure uniformity of product.

Gold in most ores is locked as small grains in
the mineral matrix as shown in this microscope
image (100 times magnification).

The standards are made in batches of one to two tonnes by milling carefully chosen ore and it is important not only that the reference value or gold concentration is accurately known, but also that it is constant through the batch. The certificate provides this reference value as well as information that may affect the use of the product.

The usual process for certifying a standard is to have it analysed by a number of high quality laboratories around the world. They do not know what values they should expect - instead each provides their best possible estimate. If a number of laboratories each give a very similar value it can be used with some confidence.

From the start, this project had some unusual features that made the use of a statistical approach particularly beneficial. Data Analysis Australia statisticians designed an innovative testing regime that sent a number of samples from each standard batch to each laboratory, balancing also for possible trends within each batch. This provided many benefits, including the ability to measure differences between laboratories while at the same time thoroughly investigating variation within each batch. The result was not only improved accuracy but also a much lower overall cost then if the batches were separately tested and certified.

As a statistician accredited both nationally and internationally, Dr John Henstridge of Data Analysis Australia took the overall responsibility for assuring quality and the reference value of these standards, a very appropriate role for a statistician.