Models for Decision Making

Article excerpt

A model can be considered as a simplified, but representative abstract of a real situation and can be either prescriptive or descriptive in nature. Prescriptive models (for example linear programming) are deterministic, often incorporating the use of algorithms (eg simplex algorithm for linear programming). Such models provide the user with the 'best' possible answer to the problem situation.

Descriptive models (such as simulation or multiple activity charts) on the other hand, merely describe the current (or proposed) situation(s) and afford the user an experimentation ability to explore different scenarios and ask 'what-if' questions of the model.


Whilst models can be considered to be either descriptive or prescriptive in nature, they can be further classified as iconic, analog conceptual or mathematical.

1 An iconic model is a physical representation of a system, and is generally constructed on a smaller scale. Physical models and photographs would fall into this category.

2 An analog model is a model where one or more of the properties of the system is represented by different properties of the model. For example a weather map showing isobars could be considered an analog model.

3 A conceptual model is used to bridge the gap between a problem and a mathematical model. In practice a conceptual model is a diagrammatical representation of the problem situation and incorporates the use of techniques such as influence diagrams and flow charts.

4 A mathematical model is one which uses formulae to describe the relationships that exist between a number of variables. Regression and time series forecasting models are examples of mathematical models.


The simplified nature of a model results in a somewhat less complicated representation than the actual situation itself. The model must, however, be as complete as possible in the relationships it assumes and describes. Overall the model must be as representative of the real process as possible. There is an obvious trade-off, when building a model, between the accuracy of the model and cost of development (including time). Any business model must, however, contain the following features:

1 Models must be fully documented and as intelligible as possible to both practitioner and third party user.

2 The model must be easy to use, maintain and, if necessary, update. Additionally, models must be responsive to changes in the business process.

3 The model should be exhaustive in its description of all the important features in the business process.

4 The model must be cost effective, is the cost of developing, validating and using the model should not outweigh the financial savings made in the business process.


In the last ten years there has been an increase in the use of personal computers and related software. Additionally, recent advances in communications provide the decision maker with increased data and information which can be easily accessed on-line.(1) PCs are now used regularly in the business sector where the spreadsheet in particular has proved to be a popular software tool.

The popularity of the spreadsheet can be explained by its natural user friendliness and because it encourages (user) experimentation. For the more experienced practitioner, the flexibility of the spreadsheet can be utilised to develop large scale formal and structured business models.


Many articles have been published which suggest that the use of modelling techniques have improved decision making effectiveness.(2)(3) Whilst this is certainly true, it is the experience of the authors that models often only improve decision making efficiency and have little or no effect upon effectiveness.

Business problems are generally complex and incorporate a degree of uncertainty. Often the models developed to describe and evaluate such problems are sizeable, involve a large amount of computation and are time consuming to undertake by hand or pocket calculator. …