The multidimensional assignment problem (MAP) is a natural extension of the well known assignment problem. A problem with s dimensions is called a SAP. The most studied NP-hard case of the MAP is the 3AP. Memetic algorithms have been proven to be the most effective technique to solve MAP. The use of powerful local search heuristics in combination with a genetic algorithm, even if it has a simple structure, provides high quality solutions without a lot of effort. We perform an experimental evaluation of a basic genetic algorithm combined with a so called dimensionwise variation heuristic and show its effectiveness against a more complex state-of-the-art memetic algorithm for MAP.