This study proposes an integrated forecasting model that incorporates both first purchase diffusion and replacement component of sales in predicting multi-product diffusion patterns. The model consists of a two-stage procedure that customers undertake during purchase occasions according to a diffusion process or replacement process, and at each occasion, they make purchase incidence decisions and product choices according to a choice model. By incorporating various exogenous factors in the choice model, the model can identify the impact of each factor on customers' purchase incidence decisions and choice decisions among various product sub-categories. This approach enables us to understand the overall process of customers' purchasing behavior and to separate total sales amount into sales to three consumer segments: first-time buyers, customers who repurchase the same product category and customers who change their previous product category. We applied the proposed model in the forecasting multi-product framework; forecasting sub-category level automobile sales and brand level mobile terminal sales in the South Korean market. The results of the empirical studies showed that incorporating the replacement components into the multi-product diffusion framework is useful not only in understanding the demand interaction among automobile categories and among mobile handset brands, but also in producing more accurate long-term sales forecasts compared to previous multi-product diffusion models. (C) 2010 Elsevier Inc. All rights reserved.