In this paper we study the expected performance of two dynamic lot-sizing heuristics (part-period-balancing and Silver-Meal) and provide an analytic framework for evaluating their expected average costs. Some recursive equations for computing the expected value of the heuristic solutions are derived and illustrated by some specific examples. A comparative study of the expected performance of the two heuristics for uniformly distributed demands is given, and the result shows that the Silver-Meal heuristic dominates the part-period-balancing heuristic in the long run.