Loyalty reward programs (LRPs), initially developed as marketing programs to enhance customer retention, have now become an important part of customer-focused business strategies. With the growth in these programs, the complexities in their management and control have also increased. One of the challenges faced by LRPs managers is that of developing models to address various forecasting issues to support short, medium, and long term planning and operational decision-making. We propose in this paper a predictive model of redemption and liability in LRPs. The proposed approach is an aggregate inventory model in which the liability of points is modeled as a stochastic process. An illustrative example is discussed as well as a real-life implementation of the methodology to facilitate use and deployment considerations in the context of a frequent flyer program, an airline industry based LRP.