The present article compares two treatments, say A and B, in the context of a clinical trial. Let X and Y be, respectively, the responses corresponding to A and B which have some continuous distributions. Here, the comparison is done through the parameter θ = P(X < Y) which is a very common functional in the context of reliability theory and survival analysis. We consider a two-stage data-dependent adaptive allocation design where the observations on X and Y are collected sequentially in two stages. We use the Mann–Whitney–Wilcoxon (MWW) statistic to provide a statistical comparison by a nonparametric approach. Various asymptotic results related to the proposed procedure are studied. The proposed procedure is also compared with the usual single-stage equal allocation design through statistical and clinical performances of the respective allocation procedures. © 2017 Informa UK Limited, trading as Taylor & Francis Group.