During an interview[1] (with one of the AirBnB founders no less), pg offered the general advice that startups should search for success using a hill climbing algorithm, and not worry about dangerous local maximums; only to be immediately confronted with the counter example of AirBnB (the cereal episode).
That company just seems to be an outlier in multiple categories.
[1] http://youtu.be/nrWavoJsEks at about 7:00