Something at the core of a lot of the systemic problems with these "sharing economy"/"p2p market" companies is that they assume workers and sellers are fungible, and are thus in something approximating perfect competition[1]. Instead, much like the dating example in the article, they are actually a Matching Market[2]: there is some close-to-optimal arrangement of buyers to sellers. Long-term happiness with the services involves finding the best fit, rather than just the first sufficient fit.

The auction system of old TR allowed for discovery of these fits via the auction mechanism. I expect that much like a lot of other econ theory mechanisms in addition to the success stories in the article, it led to a lot of large scale usability friction. In the end, it looks like TR chose a smoother system with worse individual outcomes. The holy grail then is a system that elicits information about fit without leaking too much information and being easy to use at the same time.

[1] http://en.wikipedia.org/wiki/Perfect_competition

[2] http://people.duke.edu/~aa88/articles/WorldCongressSurvey.pd...