Inventory management traditionally assumes the precise knowledge of the underlying demand distribution and a risk-neutral manager. New product introduction does not fit this framework because (i) not enough information is available to compute probabilities and (ii) managers are generally risk-averse. In this work, we analyze the value of information for two-stage inventory management in a robust optimization framework; the precise distribution of the demand is not available, but the decision-maker receives advance information on the success (high, low or moderate) of the product and knows range forecasts for the demand in each scenario. We derive closed-form expressions for the optimal order quantities and provide insights into the impact of the cost parameters. We also compute the critical probability of the baseline scenario for which the manager switches between pre-defined optimal strategies, and justify this behavior in the light of a risk-return tradeoff. Numerical experiments are very encouraging; performance improves by 10 to 50% when the manager considers three demand subregions rather than one and minimizes the worst-case cost.
Technical Report, Dept of Industrial and Systems Engineering, Lehigh University, Bethlehem, PA. May 2007.