Solving a complete problem using a local search technique generally consists of two stages. In the first, a solution to the problem is found using an initial generation technique. In this stage, normal Solver search goals (such as IloGenerate) can be used to provide initial solutions.

Assuming that a first solution with a specific cost can be found, the second stage is to apply the local search procedure to improve this solution. In its basic form, the local search procedure is carried out by making a small change to the current solution to produce a new solution. This new solution is tested against the problem constraints for feasibility, and its cost is computed. If the new solution is feasible and has reduced cost, it is accepted as the current one, otherwise the current solution remains unchanged.

This process is repeated (using different solution changes on subsequent tries) until some stopping condition is met. The stopping condition can be based on time, number of changes attempted, or the fact that no changes to the solution can be found that will be accepted under the specified conditions.

The basic local search mechanism can be made more complicated, but such complications can be left until later. For now, we illustrate the process by using local search to solve a simple problem.