Finding the best technique to identify BRAF mutations with a high sensitivity and specificity is mandatory for accurate patient selection for target therapy. BRAF mutation frequency ranges from 40 to 60% depending on melanoma clinical characteristics and detection technique used. Intertumoral heterogeneity could lead to misinterpretation of BRAF mutational status; this is especially important if testing is performed on primary specimens, when metastatic lesions are unavailable. Aim of this study was to identify the best combination of methods for detecting BRAF mutations (among peptide nucleic acid - PNA-clamping real-time PCR, immunohistochemistry and capillary sequencing) and investigate BRAF mutation heterogeneity in a series of 100 primary melanomas and a subset of 25 matched metastatic samples. Overall, we obtained a BRAF mutation frequency of 62%, based on the combination of at least two techniques. Concordance between mutation status in primary and metastatic tumor was good but not complete (67%), when agreement of at least two techniques were considered. Next generation sequencing was used to quantify the threshold of detected mutant alleles in discordant samples. Combining different methods excludes that the observed heterogeneity is technique-based. We propose an algorithm for BRAF mutation testing based on agreement between immunohistochemistry and PNA; a third molecular method could be added in case of discordance of the results. Testing the primary tumor when the metastatic sample is unavailable is a good option if at least two methods of detection are used, however the presence of intertumoral heterogeneity or the occurrence of additional primaries should be carefully considered.