The development of the models described in Repeat Buying started some forty years ago with the study of consumer panel data for particular brands showing the number of panel members making 1 or 2 or 3 and so on purchases in a year. The Negative Binomial Distribution (NBD) was found to fit this distribution well, and also that for product categories. It fitted also the distributions over time periods of other lengths.
Over some years it was realised that the average number of purchases (as opposed to volume or value) over a period was a possible measure of loyalty, and then it was observed that these averages tended to be much the same for every brand in a product category, and that this was true for many categories. This observation now seems the result of an obvious analysis, but at the time was not obvious: indeed, to many it is still not an accepted conclusion. Only for large brands over long time periods when penetration approaches 100 per cent are differences in sales volume (as measured by the number of purchases) accounted for by variations in the rate of sale.
Ehrenberg, A (2000), "Repeat Buying", Journal of Empirical Generalisations in Marketing Science, Vol. 5, No. 2