The wild potato germplasm collection of the series Acaulia, maintained at the Centre for Genetic Resources, The Netherlands, consists of 314 accessions. In previous investigations, 15 potential duplication groups with a total of 36 accessions were identified based on AFLP analysis of a limited sample per accession. In the present study, the potential redundancies, plus one additional accession, were studied with increased sample sizes to examine intra- and inter-accession variation more accurately, with the aim to reduce the size of the collection. No variation was observed within two potential duplication groups, whereas only limited differentiation among accessions was detected within seven groups, resulting in a total of 15 redundant accessions (nearly 5% of the collection). A cluster analysis of all the accessions of the collection showed that these nine groups each had a distinct identity. It was decided to maintain the accessions of the remaining six groups as separate entries based on the large differentiation observed among accessions and the absence of a clear identity. An analysis of molecular variance in the set of 37 accessions showed that 91% of the observed variation could be found among accessions. This variance component appeared unaffected when the set was analysed without the 15 redundant accessions. The invested costs to identify redundancies in the series Acaulia by AFLP analysis are estimated at k 57.3, whereas the savings achieved by reduction of the collection are estimated at k 21.0 per generation. However, a cost-benefit analysis should not only focus on the short-term return of investments, but should also consider the value of newly obtained data and information. These include taxonomic information about accessions, optimised sampling strategies, optimised regeneration procedures, additional data for core collection formation and more efficient utilisation of germplasm.