Transfer of new anticancer agents from bench to clinical trial takes in excess of 10 years and costs up to US $500 million. Despite this massive commitment, many more new agents fail in the clinical trials than are successful. The poor performance of many investigational anticancer agents in the clinic implies that the preclinical models used to evaluate them are flawed, inappropriately used or the information they generate is misinterpreted. This article reviews current practice and the range of preclinical models available. The author provides a personal perspective on what information is needed and how in the future this might best be obtained from preclinical models to more effectively inform the transfer of novel, active agents into clinical practice.