Quality & Business Improvement

Bluffer's Guide to Statistical Process Control

General notes

• Control charts - most visible sign of Statistical Process Control - first devised by Walter Shewhart of Bell Telephone Laboratories in the 1920s.
• Statistical training widespread in Japan, inspired by W Edwards Deming, disciple of Shewhart, in 1950s - rebuilding manufacturing base following World War II.
• Ford undertook major international SPC initiative in 1982.
• Suppliers may require evidence of SPC use to show good practice.

Advantages of SPC

• Reduction of costs
  (Early implementation of SPC reduces amount of unacceptable product).
• Improved reliability
  (Increased product predictability leads to closer process output to customer specification).
• Identification of special causes
  (Problems/extraordinary occurrences in the system can be graphically identified and tackled as part of continual improvement process).
• Ease of checking progress of process
  (Graphic chart format is easier to read than columns of figures).

Disadvantages of SPC

• Ambiguous message to untrained management
  An out of control process may appear at first sight to be in control to an untrained manager, who may not see the relevance, for example, points too closely together in one region of the graph.
• Perceived triviality of control chart by untrained management
  Analysis of a process may not be perceived as adding value to the company eg down time on bank service tills, types of accidents, number of errors on holiday forms.
• Possible lack of co-operation from data providers
  SPC may be perceived as an additional task on top of an already busy job; the data required may be seen as implying criticism of an employee e.g. amount of time spent processing expense forms.
• Perceived lack of support from senior management
  A junior operator may feel unsure about the financial/job implications of halting a process because of the presence of a special cause, without referring the matter upwards.

Structure of control charts

The nominal (centre line) is dotted; above and below are two solid lines representing the Upper Control Limit and the Lower Control Limit, each of which is three standard deviations from the centre line.

The limitations are that 15-25 data points are needed before the limits can be calculated and that only one characteristic can be tracked per control chart

Even if part characteristics are similar, other differences such as material type or specification limits may mean that separate control charts will have to be constructed.

Conclusion

SPC can only operate satisfactorily in a company if:

• It has the support of both senior management and relevant operators;
• Management are prepared to take immediate action in response to control charts;
• SPC presence is communciated widely throughout the company.

One way to consider SPC is to realise that while all data contains noise, only some data contain signals. If there is no way to separate the probable noise from the potential signals, then there is the probability of being misled by the noise in the data. Shewhart's Control Charts are the simplest way to separate signals from noise.