TQM - a Reminder

� � � � � � �

Management must focus on correct measurable performance attributes and navigate them. Each measurable performance attribute has both an average and variations around the average. The average of measurable performance attributes should be reduced if the attribute is undesirable (e.g. CPU time consumed) increased if the attribute is desirable (e.g. transaction throughput) or meet specifications, (i.e. be as close as possible to a pre-defined measurement). Total Quality Management (TQM) is a methodology aiming at identification of correct performance attributes, never ending reduction of variations and the navigation of performance attributes averages to the desirable position. Dr. W. Edwards Deming epitomised the TQM principals demanding the fulfilment of all or none of the 'Fourteen Points For Management' sad the elimination of the seven 'Deadly Diseases and Obstacles For Quality' (See APPENDIX No. 1).

Dr. Deming arrived in Japan after World War II and taught Japanese manufacturing managers how they could produce goods and deliver services of the highest quality faster and cheaper by following his TQM methodology.

In 1960, Dr. Deming was raised to the Second Order of the Sacred Treasure by the Emperor of Japan for his contribution to the Japanese economy. To the Japanese manufacturers, TQM become the accepted route to competitive advantage (read quality). The rigorous standards of the Deming Prize represent to all manufacturing companies a demanding guide for innovation towards quality. The quality of software products, in contrast to manufacturing products, has the reputation of not being measurable, and thus hard to be managed.

TQM, however, should be applied in the software industry, since manufacturing equipment and maintaining software have common ground, as both pump as much throughput through their facilities as possible, face demand fluctuations and constantly change the structure of their facilities. The software industry, like the manufacturing industry, should adopt the TQM approach by realising that variations in performance are inherent to every system and that clients are more concerned with variation of response time than average response time. Even though the Japanese software industry is far from matching the productivity levels of Japanese manufacturing (if it were, we would all use Japanese spread sheets and word processors), TQM, with proper adaptation and understanding, can be applied to improve the quality of software development and maintenance.

"If Japan can ... -why can not we?" Performance variations are classified either as System- atic or Special. Systematic variations are inherent to the system. The elimination of Systematic variations requires a change to the system and is fully controlled by management. Errors that can be identified cause Special variations. The elimination of Special vari- ations require a change to the way a system is used and thus, may be driven by the clients. It is management's responsibility to classify variations as Special or Systematic and act accordingly. Dr. Deming's empirical observations showed 96% of the variations to be Systematic and 4% Special. Part 5 of this paper shows how to classify variations as Special or Systematic using Control Charts. As performance variations are inherent to systems, management must aim at stabilising systems.

System stability is a desirable state of random variation. Performance variations in a stable system have the following properties: their limits are predictable, they comply to only one distribution pattern, they are systematic (i.e. due to limitations inherent in the system) and they may cause poor performance (e.g. produce too many cases of unacceptable transaction duration). A stable system delivers consistent (predictable) performance that can be improved by analysis and change. A system that is unstable produces inconsistent performance. The resulting variation may be so wide that it would be hard to say whether a given change has improved the system. We aim at having a stable system in order to predict performance, with regard to the clients' productivity, ISD's capacity planning, measure the effects of changes in the system and forecast the cost of improving clients' service. Stability is not an end in itself. Once a system is stable, work to improve quality and economy of production must commence by never ending reduction of variations and the navigation of attributes averages in a desirable direction. Given this emphasis upon the responsibility for the different types of variation. Dr. Doming was able to contain his message in a single sentence: "It is good management to reduce the vari- ation of any quality characteristic [say response time -AO], whether this characteristic be in a state of stability or not, and even when few or no defects are being produced".

� � � � � �

Hosted by www.Geocities.ws