A knowledge-based methodology for productivity analysis
P. Mohan Rao, David M. Miller · Medical Entomology and Zoology · 1988
Productivity analysis consists of measurement, interpretation, and evaluation. At the firm level, measurement is done using formal models such as profit-linked total-factor formulations. Interpretation involves assessing the numerical results of the measurement system and translating this assessment into statements such as, With 0.7 certainty, there is a serious productivity problem with direct labor. Finally, evaluation involves finding causes of the identified problems such as lack of motivation or poor training. Significant gaps exist in the productivity literature in these interpretation and evaluation areas. Moreover, there is no formal, comprehensive methodology for productivity analysis. The objective of this study, then, was to develop a formal methodology for productivity analysis using knowledge-based technologies. A spreadsheet-based system using the Profitability = Productivity + Price Recovery (PPP) model was developed and integrated into a comprehensive knowledge-based prototype system called PET. The theory and mechanics of the PPP logic are presented as a comparison with those of the American Productivity Center's (APC) model. Also discussed are the causes of performance changes indicated by the PPP and APC models: (1) resource inefficiencies, and (2) exogenous factors such as product-mix and inflexibility of resources. The utility of several analytical techniques as tools in productivity analysis was investigated. Included were linear programming (LP), statistical process control, regression analysis, and simulation. Process-control techniques were incorporated into PET. Capability requirements of a generic knowledge-based system for productivity analysis were developed. These formed the guidelines for the development of PET. This development is described--including the selection of the GoldWorks expert-system development software and the knowledge representation in PET. PET uses various user-friendly features such as screens, windows, menus, and 1-2-3-based graphics. An actual run is described in detail. PET was evaluated through experiments with a variety of scenarios and compared to established criteria for expert systems such as cost-effectiveness and the quality of its decisions. Finally, a list of possible improvements to PET and guidelines for further research are provided.