Regional flood frequency entails the pooling of data from sites within a defined region to enhance the estimation of at-site quantiles. Conventional regionalization techniques normally identify a fixed set of stations forming a contiguous region. An approach to regional flood frequency analysis that involves each site having a potentially different set of stations included for the at-site estimation of extremes was compared with a more traditional regionalization technique. The characteristics of the stations identified as being of relevance for the purposes of at-site estimation using the two approaches were contrasted and also the extreme flow values obtained were compared. The results indicated that the region of influence approach results in a group of stations with greater homogeneity than was the case for the regionalization technique and also leads to extreme flow estimates which are more accurate.