Although much is known about maximising postal survey response rates, survey clients and practitioners would benefit from a better understanding of how to maximise response for a fixed budget. Some methods are less costly than others. We present the results of a large experiment designed to test the impact of four design features on response rates and cost, using a fully-crossed factorial design: pre-notification, postcard reminder, letter design, questionnaire length. We examine costs as well as response rates which allow us to explore the extent to which omitting one or more of these features could be a false economy.