Since Solver encourages you to use the same language for the model, the prototype, and the implementation, there is not a clear line of demarcation between designing the model and implementing the program. The most that we can say about this subject is that prototyping is inherently a repetitive activity. Implementation and optimization of the implementation begin as soon as you start to improve the model that you've created.

While there may be many aspects of implementation and optimization to carry out for your overall application (for example, improving the user-interface, database connections, network transactions, etc.) we'll focus here only on implementation and optimization strictly related to Solver. We'll start by reviewing the obvious: there is no point in optimizing too soon in the development cycle. The gymnastics that you may go through to improve performance marginally in the early phases of development may very well have little or no effect on performance in the finished product while they may very well cause you real difficulty in later phases of development as you try to make necessary modifications.

Furthermore, (and this point should be equally obvious) there is no point in optimizing just any aspect of your application in wild hope of changing characteristics specific to Solver.

There are, however, practical ways to optimize with Solver.