A censored-data multiperiod inventory problem with newsvendor demand distributions

We study the stochastic multiperiod inventory problem in which demand in excess of available inventory is lost and unobserved so that demand data are censored. A Bayesian scheme is employed to dynamically update the demand distribution for the problem with storable or perishable inventory and with exogenous or endogenous price. We show that the Weibull is the only newsvendor distribution for which the optimal solution can be expressed in scalable form. Moreover, for Weibull demand the cost function is not convex in general. Nevertheless, in all but the storable case, sufficient structure can be discerned so that the optimal solution can be easily computed. Specifically, for the perishable inventory case, the optimal policy can be found by solving simple recursions, whereas the perishable case with pricing requires solutions to more complex one-step lookahead recursions. Interestingly, for the special case of exponential demand the cost function is convex, so that for the storable inventory case, the optimal policy can be found using simple one-step look-ahead recursions whereas for the perishable case the optimal policy can be expressed by exact closed-form formulas.