Nolan once stated that because location was not as big a factor for them as it is for others they could take over someone else's failed location. With a kitchen and dining area already in place the upfitting costs would be lower than starting from an empty shell. I don't know if that is the case in his prototype, but that would explain the lower build out cost. With restaraunts regularly abandoning lease space as their early rent concessions run out, there is a pretty endless supply of already upfit spaces for those brave enough to take them in locations that are a little marginal.
Market Data 
Markets 
AI
