| Order promising and order fulfillment capabilities have become a strategic differentiator in the make-to-order e-business environment. Conventional available-to-promise (ATP) models keep track of the uncommitted portion of existing finished goods inventory and planned production quantities, and promise customer orders based on this static push-based availability database. In contrast, advanced optimization-based ATP models use dynamic resource allocation engines to directly match resources, including material and capacity, with customer orders.; To study ATP issues, we introduce a push-pull framework as the foundation of this dissertation. Under the push-pull framework, we define three fundamental types of ATP models: push-based ATP, pull-based ATP, and push-pull integrated ATP, and further classify order promising practices into three categories: push-dominated ATP, pull-dominated ATP, and push-pull integrated ATP.; Pull-based mixed integer programming models are developed to model two popular customer order formats based on real-world business scenarios and data. These models take into account a variety of supply chain constraints, such as material compatibility and substitution preferences. We report on several simulation experiments using the ATP models to investigate various system design and business policy issues. A push-based stochastic revenue management model, based on a dynamic resource allocation policy, is also established. The analytic model optimizes the total profit over four commonly-seen demand classes for non-perishable products by properly reserving resources between current and future demand stages. Sensitivity analysis from numerical experiments reveals interesting and important insights for improving resource allocation efficiency needed in real-time order promising.; In addition, this dissertation studies push-pull integration issues. We create a goal programming model and a stochastic programming model that coordinate push and pull mechanisms. The first model relies on inventory control parameters to enforce a resource reserve policy. The second model lets current actual customer orders compete with future pseudo customer orders in multiple scenarios. Simulation experiments are conducted to test these models. We also provide an integration architecture that gives a framework for approaches of this type. |
