Modelling the Advertising Effects - Part II (Optional)

In this segment, you will learn how to model the market share of a target brand as a function of its share of marketing effect in the total marketing effort done by brands of a similar category or brand value. 

So, now you have an equation that looks quite complex. But how would you solve this regression model? The idea is similar to what you saw in the multiplicative model earlier, i.e., using log transformation.

There are three major advantages of using the multinomial logit model. These are as follows:

1. This model includes the 'competitive context' since the sum of the market shares of all brands is always 1, and the market share of any brand only can lie between 0 and 1. So, this becomes an optimisation question with range constraints. The market share of the target brand is always considered in the presence of all rival brands.

2. This model rightly estimates the S-curve of the market share given the independent factors. In the case of advertising, this implies that the market share is low at a very low level of advertising, and then increases rapidly beyond a certain inflection point, and flattens again out at a very high level of advertising.

3. Since the multinomial logit model has an inherent S-shaped curve, the elasticity of the independent variables shows a bell-shaped normal curve. This relationship means that at a very high level of marketing effort, each percentage increase in the marketing effort translates into a very small rise in the market share. The same holds true at a very low level as well. It attains the highest responsiveness somewhere at the middle level.