A sequential approximate optimization method is used to optimize computational expensive or non-smooth output behavior of simulation models. In this paper a flexible and compact object-oriented framework is proposed that supports the implementation and use of a sequential approximate optimization strategy. The framework distinguishes data objects and computational functions. The data objects relate to the storage, flow, and control of information that is generated during the sequential approximate optimization process. The computational functions represent ‘blackbox’ numerical routines with fixed input and output (not necessarily object-oriented), e.g. from external software libraries. They carry out specific computational tasks, such as the determination of a design of experiments, a linear regression analysis, or the solution of a nonlinear-programming problem. A set of data object classes has been defined. Basic and user defined classes can be distinguished. The basic classes represent the ones that are needed in any sequential approximate optimization strategy. The user defined classes include the SAO problem specific classes. The user may select built-in library classes or implement new classes using the SAO class templates available. The methods of the SAO classes are interfaced with the (external black-box) computational functions.