Many crop genetic resources collections have been established without a clearly defined conservation goal or mandate, which has resulted in collections of considerable size, unbalanced composition and high levels of duplication. Attempts to improve the composition of collections are hampered by the fact that conceptual views to optimize collection composition are very rare. An optimization strategy is proposed herein, which largely builds on the concepts of core collection and core selection. The proposed strategy relies on hierarchically structuring the crop gene pool and assigning a relative importance to each of its different components. Comparison of the resulting optimized distribution of the number of accessions with the actual distribution allows identification of under- and over-representation within a collection. Application of this strategy is illustrated by an example using potato. The proposed optimization strategy is applicable not only to individual genebanks, but also to consortia of cooperating genebanks, which makes it relevant for ongoing activities within projects that aim at sharing responsibilities among institutions on the basis of rational conservation, such as a European genebank integrated system and the global cacao genetic resources network CacaoNet.