By testing targeted individuals in a clade. Basically, if you know the structure of a haplogroup tree beforehand, and you find an SNP in an individual sample, you work backwards along the tree (or forwards, or whatever way is most efficient) until you reach a node where all the individuals downstream of that node test positive for the SNP, and all those upstream test negative. If another SNP is already at that location, then the existing SNP is phyloequivalent to the SNP you've found.