As I explained above it is both part of a process of elimination of matches with similar surnames who do and do not match at R-L21, in this case, coupled with a comparison of results with surname matches at FTDNA who also match at R-L21 and have done deeper SNP testing or BIg Y testing and have a terminal SNP further down their tree which represents a mutation that is more recent than R-L21. Those results are published in most surname DNA projects at FTDNA. There are several ways to easily determine if R-L21, in this case, is upstream of a marker further down the tree. I use the free tools at GeneticHomeland.com. If a tester's terminal SNP matches another tester's terminal SNP far enough down the SNP tree, they have a Y DNA match at some level, potentially in a genealogically meaningful time frame, based on when the mutation which created their known terminal SNP occurred. The 37 marker test at FTDNA costs $169 and an SNP test at YSEQ costs $23 the first time, since you have to submit a DNA sample, and $18 thereafter. YSEQ will also create a new SNP test for free. Would I recommend that someone using this method also take a Y DNA test? It depends on what kind of primary record evidence exists to link the two shared surname lines, the level of SNP testing which was done at FTDNA, their financial situation, how anxious they are for answers, etc. I might recommend that the tester just observe and wait for new terminal SNP results for their match at FTDNA to be published and then do another YSEQ test to see if they still have a match. I'm not saying STRs and SNPs are the same thing but that they can be used in the same way.
I have paid for a several other folks' Y DNA tests at Family Tree DNA in order to prove out my line. Now, when I find a new potential surname cousin, instead of paying $169 for a 37 marker test, which is the lowest marker test that I would recommend (I do have some 25 marker surname matches which fall away at 37 markers), or try to get them to pay for it, I suggest that they take a $23 test at YSEQ and see if they match my terminal SNP. If they do match, share my surname and share some basic data in common, I may call that good, or I may recommend a Y DNA test. It depends on the specific situation. If they don't match then our shared surname is a coincidence.
I also don't usually recommend that more than one member of a surname cousin group, which is already proved by Y DNA testing, take a Big Y test, SNP packs or deeper SNP testing, unless it is to verify that they continue to share a terminal SNP. That may vary, depending on specific circumstances. FTDNA does not always have individual tests available for specific terminal SNPs, so it is usually cheaper to just do the testing for that single SNP at YSEQ. The downside of that is that it doesn't get published on FTDNA's results pages automatically.
My surname cousin group is a good example of how this works. Four members have done 111 markers, 15 67 markers, 5 have done only 37 markers, one Big Y, one SNP packs and three have done terminal SNP testing at YSEQ. We share an MRCA who was born about 400 years ago. My group does originate from Virginia and some burned records counties make if uncertain exactly how we all connect, so we may need more data at some point. We share a terminal SNP, which, so far, is only represented by my surname family group. That may change in the future, but so will our terminal SNP as more mutations are discovered.
Does this make sense?