My estimate: N people are talking about an email client -> there will be ~ log2(N) crucial features that divide the group into large subsets
