It doesn't seem incompatible to me at all. As a general rule, the parts of your staff that scale O(n) with revenue should be cost sensitive but the parts that

It doesn't seem incompatible to me at all. As a general rule, the parts of your staff that scale O(n) with revenue should be cost sensitive but the parts that scale O(log n) or O(1) should be quality sensitive. As long as they can keep their growth rate up, the ratio of their staffing expenses to revenue should go down over time, regardless of how high the original denominator was. Thus, if you're a high growth potential company (which, if you compare the size of PLoS to Elsevier, seems plausible), you should pick the city that maximizes your growth potential. Since traditional publishing was such an intrinsically distributed process in the first place, PLoS has an unusually easy time building a distributed workforce for all of its journal specific workforce like editors which keeps those costs low. At the same time, the team working on functions that span all journals like software or marketing can centralize in a city like SF which bats above it's weight in those areas.