Ed Glaeser provides a decent description for why these guys just won a bunch of money.
Short version: as with many things, when you move away from Econ 101 models to add legitimate complications, lots of results change. This work is basically about adding search and matching into labor market models. Firms and workers are both looking for a perfect match, but will settle for an imperfect match as at some point that's better than continuing the search.
