On FlightCar, I would suggest: First, we have to notice the nature of the customer 'load', e.g., on their shuttle car/bus, front desk, rental operation, etc.

On FlightCar, I would suggest: First, we have to notice the nature of the customer 'load', e.g., on their shuttle car/bus, front desk, rental operation, etc. That is, this load has peaks and valleys, some fairly predictable based on airline schedules, some based on weather, and some just 'random' things independent of anything in the past available to be used as a predictor. Second, for such a load, there nevertheless will be a strong effect from the law of large numbers that as the load increases the standard deviation of the load will become a smaller and smaller fraction of the average load. That is, a large load is much more predictable than a small load. Third, then, for efficiency in staffing, shuttle cars/buses, happy customers, etc. want a large load. Fourth, notice that the customers who leave their cars are likely residents of the area surrounding the airport. I.e., if Joe lives in Kansas City and flies to Los Angeles, maybe that means he wants to leave his car at the Kansas City location, that is, use the service in Kansas City, but does not much mean he wants to use it in Los Angeles. Fifth, notice that a location of FlightCar in Kansas City is not in much competition with a competitor in, say, Cleveland. That is, this business is heavily geographically local. So, net, for a large load, for repeat customers at a location, and to defend against competition at a location, want to concentrate on that location and do well there before moving to another location. So, this business can have a fairly strong geographical barrier to entry. That is, it is basically a local business. So, a 'big national or international rollout' should be delayed in favor of just doing well one city at a time.