<p>Ex situ conservation breeding programs can provide essential support for recovering endangered species. However, breeding recommendations and strategies may be suboptimal due to untested or incorrect assumptions about conservation units, hybrid ancestry, inbreeding history and family relationships. While genomic approaches have been used to address specific questions in ex situ programs, they are not yet routinely used to test and adjust assumptions, often due to limitations in sample availability and quality particularly at the start of a new program. Using a new conservation breeding program for the endangered red siskin (<i>Spinus cucullatus</i>) as an example, we demonstrate how genomic data from a relatively small number of samples, of uneven quality, may be employed to critically examine fundamental assumptions. Using genetic data from 72 individuals, we confirmed strong differentiation between birds from Venezuela and Guyana, supporting an assumption that they should not be mixed in a conservation breeding program. Molecular estimates of genetic diversity and inbreeding also aligned with assumptions that widespread historical illegal trade had introduced the species’ Venezuelan wild diversity into U.S. avicultural flocks. We found no evidence of hybrid ancestry with canaries, which would have resulted from intentional crosses to produce red-colored canaries. Finally, while observational pedigree data were generally reliable, molecular relationship assignments and kinship estimates had high accuracy only for first-degree relatives, limiting our ability to evaluate assumptions about relatedness. Our results highlight the utility of constrained but informative datasets for developing conservation breeding programs, supporting effective management for the long-term survival of endangered species.</p>