Predicting the long tail of book sales: Unearthing the power-law exponent |
| |
Authors: | Trevor Fenner George Loizou |
| |
Institution: | Department of Computer Science and Information Systems, Birkbeck, University of London, London WC1E 7HX, UK |
| |
Abstract: | The concept of the long tail has recently been used to explain the phenomenon in e-commerce where the total volume of sales of the items in the tail is comparable to that of the most popular items. In the case of online book sales, the proportion of tail sales has been estimated using regression techniques on the assumption that the data obeys a power-law distribution. Here we propose a different technique for estimation based on a generative model of book sales that results in an asymptotic power-law distribution of sales, but which does not suffer from the problems related to power-law regression techniques. We show that the proportion of tail sales predicted is very sensitive to the estimated power-law exponent. In particular, if we assume that the power-law exponent of the cumulative distribution is closer to 1.1 rather than to 1.2 (estimates published in 2003, calculated using regression by two groups of researchers), then our computations suggest that the tail sales of Amazon.com, rather than being 40% as estimated by Brynjolfsson, Hu and Smith in 2003, are actually closer to 20%, the proportion estimated by its CEO. |
| |
Keywords: | Long tail Power-law distribution Stochastic model |
本文献已被 ScienceDirect 等数据库收录! |
|