AbstractThe relative performance of several sequential methods of planning multiple delivery journeys has been tested on batches of problems which varied in size, journey restrictions, delivery pattern and depot location.The saving criterion, of the methods tested, produced the best results for the conditions tested and required less than 0·1 min CDC 6600 central processor time for a 400 delivery problem; the time required varied asN1·6whereNis the number of deliveries.The relative performance of different methods varied considerably both between problems conforming to the same specifications and between batches of problems with different specifications.To compare the performance of different methods for any but extremely limited purposes, samples of results for a variety of conditions are needed; even so, in the absence of an explanation for the performance variations, only tentative conclusions are possible.