Strategic Modelling and Business Dynamics: A Feedback Systems Approach

10.7. Tests of Learning from Simulation

Simulations can achieve more than mimicking history. Like Papert's gears of childhood, a simulator is a transitional object to improve understanding through testing and reflection. With a simulator it is possible to conduct a wide range of experiments to better interpret past behaviour (as in the base case above), to explain surprise behaviour, to learn about future behaviour, to test pet theories and hypotheses and to correct faulty expectations. Whenever a model provides useful new insight (to the client or to the modeller) then it passes the test of learning from simulation. Some of the most common learning tests are described below.

10.7.1. Comparing Simulations with Expectations and Interpreting Surprise Behaviour

How will the future unfold if no changes are made to the conduct of the business? Managers have their own views and opinions on this question against which to compare simulations and invite discussion. The soap industry model was run five years into the future with the same parameters as in the base case to reflect the assumption that the firm's decision-making processes continue unchanged. This business-as-usual projection led to several insights deemed worthy of discussion. It showed that the retail price of bar soap will rise in the future. At first glance, this result is surprising because the bar soap business is price sensitive and highly competitive. On reflection there is a good explanation. The price rise is a response ...