We are concerned here with the ordinary private consumer's purchases of non-durable consumer goods. These goods are usually characterised by being marketed in pre-packed and branded form. Data about such purchases are obtained by market research techniques such as, in the case of Attwood's, the continuous consumer panels (see reference 1) based on random samples of either households or individuals and operated in various European countries including Great Britain. (The samples used in market research are almost always large in the statistical sense, so that no small-sample theory is required.) The basic unit of time for measuring consumer purchases is usually a week, one week being generally like another. Most analyses are, however, made over periods of 4 or 13 weeks. For any such period of time, we therefore know how many consumers in the sample bought 0, 1, 2, 3, 4, or, in general, r, units of the given product, i.e. we know the frequency distribution of purchases. We generally also know what each of the consumers bought in preceding periods and can continue to watch his subsequent purchases. The problem considered in this paper is the fit of the negative binomial distribution to such data. Product-fields analysed include the following: Bread, Breakfast Cereals, Canned Vegetables, Cat and Dog Foods, Cocoa, Coffee, Confectionery, Detergents, Disinfectants, Edible Fats, Food Drinks, Household and Toilet Soaps, Jams and Marmalade, Polishes, Processed Cheese, Sausages, Shampoos, Soft Drinks, and Soups.