No Title

3 A pilot model

The pilot model was applied to four brands of a particular type of alcoholic beverage. Each brand has two sizes which were considered. The statistical procedures described in Section 1 was applied to 96 weekly observations of price and volume (number of bottles sold) for each of these eight brand-sizes. A graph of the values of the local regression coefficients is given in Figure 1. Three brands/sizes had already been eliminated by the application of the linear analysis including both sizes of one brand.

We see that round about date 30, apparently significant changes in the local regression coefficients began to occur and that they remained unsettled until about date 60. these changes were picked up by the model and reported (in a still fairly crude transcript window). the report, with brand names suitably disguised, was:

Having picked up the dates at which the coefficients became unsettled, it went on to note:

It then provided similar information for the remainder of the dates in that sequence.

The reasons for these dates to have corresponded to unsettled behaviour of the local regression coefficients is clearly a matter of domain knowledge. What we do know from the domain experts (but did not know when conducting these experiments) was that a change (some would say a degradation) in the formulation of the focus brand was introduced at about date 30. As we can see in Figure 1, the coefficient on the focus brand's own price declined sharply from that time. Moreover, Brand F2, which is the larger size but same brand name as the focus brand, became more competitive with the focus brand itself in the sense that the market share of the focus brand became more sensitive to the price of the larger size. During the unsettled phase, brands B1 and B2 became more competitive with the focus brand for a while. These are two sizes of a rather (but not the most) down-market brand name.

These changes in the competitive structure of the market were captured by IMIS in the following way:

Here, IMIS is supporting one side of a discussion among the domain experts which followed the formula change. Some argued that market share was lost to brand A while others argued that share was lost to brands B1 and B2. The conclusion of IMIS is that the focus brand became more competitive with larger sizes of its own and relatively down-market brand names.

Finally, IMIS eliminated Brand A from the competitive set on the grounds that brand B2 had taken some of its competitive position with respect to the focus brand and that Brand A had simply become less competitive over the observation period. These two statements together imply movement of the focus brand rather than brand A in the overall competitive space.

No Title - 23 AUG 96

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

Generated with CERN WebMaker