In this section, we aim at answering our first question: How does the optimal assignment of bikes to stations depend on the probability distribution of the rental demand? Table 3 shows the optimal number of bikes x * i , i.e. the number of bikes assigned to each bike-station i ∈ Bat the beginning of the service, under the four considered probability distributions, using 500 scenarios. The results show that the optimal solutions under the assumption of Uniform, Normal and Log-normal distributions are similar, with a total number ranging from 200 to 205 (see last column), while a different behavior is obtained with the Exponential distribution: When we place more under the former distributions, we place less under the latter, and vice versa. This is mainly due to the different shape of the Exponential distribution. The average op- timal number of bikes over all the bike-sharing stations is 9 under the assumptions of Uniform, Normal and Log-normal distributions, while it is 16 bikes under the assumption of the Exponential distribution. Interestingly, the optimal stochastic solution suggests a fleet size, as the total number of bikes initially allocated to the stations, to be less than 1/3 of the 660 = 30 ·22 available slots, for Uniform, Normal and Log-normal distributions while more than 1/2 for the Exponential. This low “utilization rate” of slots is there to avoid too high overflow, stock-out and transshipping costs.