W e discuss the development and implementation of CHAN4CAST, a sales forecasting model, by pack size, category, channel, region, customer account and a Web-based decision support system (DSS) for consumer packaged goods. In addition to capturing the effects of such variables as past sales, trend, own and competitor prices and promotional variables, and seasonality, the model accounts for the effects of temperature, significant holidays, new product introductions, trading day corrections, and adjustments to the wholesale level. In general, the model forecasts sales volume satisfactorily for a leading consumer packaged goods company. The DSS enables top-and mid-level executives in sales, marketing, strategic planning, and finance to develop accurate forecasts of sales volume, plan prices, and promotional activities over a long time horizon; to track sales response to marketing actions over time; and to simulate forecast scenarios based on possible marketing decisions and other variables. CHAN4CAST is being rolled out for more users and more divisions in the company. The key take-aways are that successful development and implementation of a rigorous marketing science model require a strong internal champion, a careful balance between modeling sophistication and practical relevance, good diagnostic features, regular validations, and greater attention to the development of a fast and responsive DSS.

No abstract

Category management-a relatively new function in marketing-involves large-scale, real-time forecasting of multiple data series in complex environments. In this paper, we illustrate how Bayesian Vector Autoregression (BVAR) fulfils the category manager's decision-support requirements by providing accurate forecasts of a category's state variables (prices, volumes and advertising levels), incorporating management interventions (merchandising events such as end-aisle displays), and revealing competitive dynamics through impulse response analyses. Using 124 weeks of point-of-sale scanner data comprising 31 variables for four brands, we compare the out-of-sample forecasts from BVAR to forecasts from exponential smoothing, univariate and multivariate Box-Jenkins transfer function analyses, and multivariate ARMA models. Theil U's indicate that BVAR forecasts are superior to those from alternate approaches. In large-scale forecasting applications, BVAR's ease of identification and parsimonious use of degrees of freedom are particularly valuable. KEY WORDS Beyesian vector autoregression;multivariate time series modelling; competitive dynamics; category management; dynamic conditional forecasts; state-space modelsIn the past few years., the single-source industry has become a dominant force in packaged goods marketing research. Single-source (scanner) data are a database of electronically collected information (using supermarket checkout scanners) on household purchases over a period of time. This database also contains in-store information on item prices, promotions, and advertising and is available at the household, store, or market level. Single-source data permit

No abstract