Abstract:

The present paper describes the devices adopted in order to construct the transition matrix which is going to serve as the model's input : they involve breaking down each purchasing sequence into pairs of elementary purchases. But this in turn created a new series of difficulties : the aggregating of unequal intervals between purchases, the impossibility of retracing an individual loyalty to a brand from repurchasing rates. The proposed solution consists of segmenting the market according to two dimensions that emerge from the Markov mechanism -purchasing intensity, and loyalty which also prove very instructive as far as marketing is concerned . The experiments carried out on two non-durable consumer product markets consisted of comparing predicted shares with actual shares over a period of three years, and enabled us to define the degree of reliability of the model.