For a black-box global optimization algorithm to be truly global, some effort must be allocated to global search, that is, search done primarily to ensure that potentially good parts of the space are not overlooked. On the other hand, to be efficient, some effort must also be placed on local search near the current best solution. Most algorithms either move progressively from global to local search (e.g., simulated annealing) or combine a fundamentally global method with a fundamentally local method (e.g., multistart, tunneling). DIRECT introduces a new approach: in each iteration several search points are computed using all possible weights on local versus global search (how this is done will be made clear shortly). This approach eliminates the need for ‘tuning parameters’ that set the balance between local and global search, resulting in an algorithm that is robust and easy-to-use.