The paper considers the problem of modeling the lead-time demand for the multiple slow-moving inventory items in the case when the available demand history is very short, and a large percentage of items has only zero records. The Bayesian approach is used to overcome the mentioned problems with the past demand data: it is proposed to use the beta-binomial model to predict the lead-time demand probability distribution for each item. Further, an extension of this model is developed that allows accounting for the prior information regarding the maximum expected probability of demand per period. Parameter estimation and Bayesian forecasting routines are derived for the new model. The efficiency and practical significance of the obtained results is proved by the simulation study.