A multistage stochastic optimization model is presented to address the scheduling of supply chains with embedded multipurpose batch chemical plants under demand uncertainty. In order to overcome the numerical difficulties associated with the resulting large-scale stochastic mixed-integer-linear-programming (MILP) problem, an approximation strategy comprising two steps, and based on the resolution of a set of deterministic and two-stage stochastic models is presented. The performance of the proposed strategy regarding computation time and optimality gap is studied through comparison with other traditional approaches that address optimization under uncertainty. Results indicate that the proposed strategy provides better solutions than stand-alone two-stage stochastic programming and two-stage shrinking-horizon algorithms for similar computational efforts and incurs much lower computation times than the rigorous multistage stochastic model. (c) 2006 American Institute of Chemical Engineers AIChE J, 52: 3864-3881, 2006.