A novel technique for empirical optimisation is presented, called dynamic local search algorithm (DLS). The search algorithm starts exploring the solution space only along one of its dimensions at any one time. This is done by perturbing this variable randomly along opposite directions of that dimension, creating two more variables. The magnitude of the perturbation is designed to explore Local and global areas. If these perturbations were found to be minimising (or maximising) the function, then the original variable is moved toward them in a gradual manner. This gradual move will prevent the system from getting trapped in local minima. Repeated operations of the above process will gradually guide the system towards the global minimum of the function. The results arising from the application of DLS to the De Jong's test functions compare positively with those from other global and local methods.