This editorial discusses an illustration of the potential hindrances to the diffusion of modern methodologies in the practitioners' (i.e. the buyers of research, not the consultants) community. Taking the example of classical regression analysis based on store-level scanner data, the authors discuss the potential limitations of the classical regression model, with the example of the occurrence of 'wrong' signs and of coefficients with unexpected magnitudes. In an interview with one of the authors, a (randomly picked) Senior Marketing Research Manager at a leading firm of packaged goods reports his/her experience with econometric models. To him/her, econometric models are presented as a 'black box' (his/her written words). In his/her experience, they provided results that were 'quite good' in a 'much focused' context only. There were experimental data obtained with a Latin square design and the analysis included a single brand with only four stock-keeping units (SKUs). The company 'dropped' the more 'ambitious' studies, which analysed the effect of the retail promotions run by all the actors in a market because of a lack of predictive accuracy (his/her written words are in quotes). The authors suggest that Bayesian methodology can help open the black box and obtain more acceptable results than those obtained at present. Copyright.

Additional Metadata
Keywords Bayesian analysis, Forecasts, Latin square design, Regression analysis
Persistent URL,
Journal Applied Stochastic Models in Business and Industry
Bemmaor, A.C, & Franses, Ph.H.B.F. (2005). Editorial - The diffusion of marketing science in the practitioners' community: Opening the black box. Applied Stochastic Models in Business and Industry, 21(4-5), 289–301. doi:10.1002/asmb.553