A vital component of conservation is to sample germplasm from wild populations for safeguarding ex situ (e.g., in seed banks, though the concept also applies to animals via cryopreservation or zoos). For decades, conservationists have commonly heeded a logically sound, but perhaps suboptimal, minimum sampling guideline- to sample from 50 individuals per population. Here, I demonstrate how sampling can be improved based on two considerations that are neglected in this common, simple guideline, and have not previously been tested. First, I consider a fundamental aspect of population biology- sharing of genetic material among populations through migration. Second, I consider a fundamental aspect of ex situ collections maintenance- loss of plants through germination failure, disease, and active use (e.g., research, seed exchanges). I first simulate metapopulations with a wide range of migration rates, population sizes, numbers of populations and demographic (i.e., bottleneck) histories. Then I determine minimum sampling to preserve a sufficient number of allele copies to account for various degrees of collection attrition. My results show that sampling seed from approximately 200 to 300 individuals in total across a species' geographic range may suffice for a wide range of plant population systems, if all seeds germinate and produce plants that survive in perpetuity. However, to compensate for expected losses over time, sampling should be increased by a factor closely related to the expected loss rate, meaning that a robust minimum collection for ex situ gene conservation will often be 1000 individuals or more. More work is needed to establish final guidelines, but I do provide a summary table for practitioners. I conclude that seed collections planning must consider the plant's biology, collection maintenance, and desired characteristics of the collection such as the number of allele copies and the type of allele targeted. These results emphasize a need for thoughtful deliberation and decisions by the curator and collector, and renewed discussion of conservation targets for ex situ collections to remain viable over hundreds to thousands of years.