Background. Diagnostic confirmation of spindle-cell melanoma (SM) or desmoplastic melanoma (DM) as a melanoma can be challenging. In conventional melanoma (CM), a recently established fluorescence in situ hybridization (FISH) assay for RREB1, MYB, CCND1 can be helpful. Here, we determined the presence of RREB1, MYB, and CCND1 abnormalities in an SM/DM/mixed cohort. Methods. We assembled 49 cases and performed 3 separate hybridizations for RREB1/MYB/CCND1. We assessed clinical utility in diagnostically challenging cases and performed a cost and turnaround time analysis. Results. With regard to the diagnosis of melanoma, the FISH assay is 76% sensitive (n = 31/41 true positives melanomas) and 88% specific (n = 1/8 false positive desmoplastic nevi). The prevalence of abnormalities in DM is lower (12/19 cases, 63%; P = .03) than in SM (15/18 cases, 83%; P = .27), mixed (4 of 4 cases), or the reported sensitivity in CM (345/411 cases, 84%). The implied genetic differences in DM result in a higher false negative rate in DM (37%). Despite these limitations, when restricted to diagnostically challenging cases (n = 23), the FISH assay and, in particular, RREB1 was able to confirm melanoma in 70% (n = 16/23). Individual probe sensitivities (RREB1 > MYB > CCND1) and a cost and turnaround time analysis argues for a 2-step test algorithm that reduces the economic impact of FISH testing considerably (~55%; n = 69 vs 123 hybridizations). Conclusion. We propose a step-by-step genetic testing algorithm to support the diagnosis of melanoma in the setting of SM/DM and show that FISH testing is useful in diagnostically challenging cases.