No Title

5 The demand side and model validation

In part, the validation of the models generated by IMIS will turn on their plausibility in the eyes of the domain experts. However, another part of the IMIS project deals with the demand side and the consumers' preferences for different characteristics for different purposes. An early specification of these context-dependent attribute preference states (CDAPS) was reported by Moss and Edmonds (1996). In that paper, we showed that CDAPS could be parameterized on the basis of a small subset of the data and explain them within an IMIS-type model and track observed market shares better than OLS models in the sense that the IMIS- generated shares exhibited a lower mean absolute percentage error and root mean-squared error outside the sample period.

It was assumed in that paper that the perceived characteristics of the various brands were constant over the observation period. However, a useful finding from the IMIS models reported here is that either characteristics or preferences do change within the two to three years covered by the data in each case. Presuming that culture changes more slowly than perceptions of brand characteristics, it is natural to use results such as those obtained here to get from the domain experts their views of how brand perceptions have changed and then to make sure that these views are consistent with IMIS's interpretation of the data. That is, the domain experts' accounts should tally with the changes identified in the transcript fragments reported here. It will then be possible to assess whether we can still track the share data using the time pattern of characteristic perceptions together with the specifications of the CDAPS. If we can, then we shall have a complete and consistent account of the competitive structure and how it changed during the period for which we have data.

No Title - 23 AUG 96

[Next] [Previous] [Top] [Contents]

Generated with CERN WebMaker