Looking at Figure 1, it is not a good plot since customers' numbers are really small and hard to visualize in the plot. Thus, I normalize the distribution of subscribers, and that of customers, i.e. devide them by the total number of rides for that user type.