In this paper we describe an iterative development process to deal with speculative constraint optimization projects. Speculative projects are ill-defined in nature, as they are often new to the customer organization who wishes to anticipate market changes, but also because their main complexity lies in completing the problem definition. We consider one speculative application tackled in the Chic-2 Esprit project by different partners: Risk management in the area of energy trading 1. The ill-defined and incomplete components of the problem are: 1) model adequacy, 2) multi-criteria objective, and 3) forecast data. We show how the constraint technology can help refining the problem definition by building iterative models and methods to address the incomplete parts incrementally. Each partner applied his own technology using constraint programming, mathematical programming and local search techniques. Finally we discuss the new potentials for constraint technology to deal with stochastic problems.