To define different human races, people typically use genetic distances and the variation in phenotype between groups.
However, cladistics have become popular among evolutionary biologists and for good reason. For the uninitiated, cladistics are quite simple; groups are related insofar as they have a common ancestor, to the degree of how recent that ancestor lived. I’ll quote myself from the comments section of a Pumpkin Person post on the Big Three racial groups:
To define human races you could use cladistics, genetics, or phenotype.
From a standpoint of cladistics, Australoids, Bushmen, Caucasoids, and Mongoloids are all monophyletic groups with a set of common ancestors that are exclusive to their own kind. The same is probably true for pygmies and Hadza. The remaining African Negroids are a catch-all for everybody else: they have no common ancestors shared with one another that are not shared with Europeans, Asians etc. The only reasons that Congoid African groups resemble one another more than they resemble Europeans or Asians is that they had similar selective pressures and that they interbred.
I’m leaving admixture out of this. Australian Aboriginals for example are share some genes with Europeans, but the Aboriginal racial group has a founding population separate from Europeans. Just because two streams are reunited, doesn’t mean they didn’t split.
Next I compare the cladistic model of categorizing human races with the genetic distance model:
From a standpoint of genetics (most of which may be neutral genome as you suggest) the biggest autosomal genetic distances are between Africans and non-Africans, Caucasians and Asians, and between various African groups (http://www.pnas.org/content/108/13/5154/T1.expansion.html) such that pygmies, Hadza, Bushmen etc are roughly as genetically distinct from one another and from Congoids as Europeans are from Asians. So that lines up with the cladistics. These data would be clearer if someone cross referenced them with admixture analyses: the genetic difference between San and Bantu peoples for example would be even larger if you accounted for the fact that there has been admixture. That said, homozygosity can inflate these distances.
The reason that these two line up is that genetic distances are used to determine, say, whether and when the Bushmen split from other Africans. They’re two different models of the same reality, and sometimes the same data.
This makes it all sound useless: just look at the genetic data, you may insist.
But it adds perspective. Genetics tell us that the Micronesians and Africans are very different, but judging by phenotype you would consider them similar. Cladistic models built with genetic data tell us that Micronesians share a set of ancestors with all non-Africans that they do not share with Africans; they may look similar and act similar, but they’re on a different branch of the family tree.
The question ultimately comes down to whether you’re more concerned with phenotypes or with evolutionary relationships.