Multisite Sharing: Searching the Golden Rule

7 Apr 2022    |    4 min read


Sharing mobility is a whole world. 

Starting a new service of sharing mobility is tough.

You have to choose what kind of fleet you want (mopeds, kicks, scooters, bikes, cars… or a mix of them). 

You have to choose what kind of the selected type of vehicles you want (NIU, SuperSoco, and so on).

You have to choose a provider for the connectivity of your fleet

You have to choose the name of the service, its colors, the app. 

You have to choose. A lot. 

Even when everything above is set, you have to answer the million-dollar question. How many vehicles do you need for your service? 

The decision is crucial to the fortunes of your service, and still, there is no clear answer. 


Because they will tell you that every city is different. Every country is different. Every person has habits, rules, timetables. 

Well, guess what. 

It’s a lie, and this article is all about this. 

From our experience as 2hire, we found a rule which helps you to answer the most difficult question you will ever face in the starting process of a mobility service: how you choose the number of vehicles in your fleet.

The Analysis

Finding a rule that fits different cities, with different climates, and different habits is not an easy process. In the summer of 2021, however, at 2hire, we had the opportunity to observe a single service act across multiple cities. The Elerent service has been active in almost 30 cities in Italy, from North to South, from East to West, from the mountains of Abruzzo to the seaside resorts of Sicily, Sardinia, and Puglia. The dataset lends itself to analysis for many reasons, but one of the things that makes it so valuable is that each site uses the same app, with similar costs for end-users. Often, when comparing different services, too much diversity in user experience can lead to lost value in the analysis. This is not the case here. 

For each city, for each day in August, the following variables were selected as variables: 

  • TripID: unique trip identifier 
  • VehicleID: the unique identifier of the vehicle that made the trip
  • UserID: the unique identifier of the user who made the trip 
  • Price: the price paid by the user for the trip. 

At the end of the month, for each city, the information collected contained: 

  • Revenues
  • Fleet size
  • Number of active users, i.e. with at least one trip 
  • Number of trips

As expected, these numbers are not comparable with each other: it would be wrong to compare the earnings of a city with a fleet of 60 vehicles with those of a fleet of 10. The same goes for the number of trips, users, or active vehicles. What is needed is a different measure, generated by putting together more than one of the metrics above. 

Our analysis considered the Net Profit per Trip (NPT), calculated as the algebraic sum of profits, minus some fixed costs such as operational costs, IoT provider, and so on. 

A sharing service will have to sustain other costs, but they will generally be costs that will impact with the same proportion all the earnings and that, for now, we can exclude. 

We put the four standard quantities (revenues, fleet size, number of active users, and number of trips) in a correlation matrix with the Net Profit per Trip. The goal is to understand if the NPT is valid as a control metric. A correlation matrix is a tool for understanding whether quantities are in any way proportional to each other. The higher the correlation between two variables, the closer to 1 is the value. If two metrics are not correlated, the corresponding number on the matrix tends to be -1.

Fig. 1 Correlation matrix

As you can see, the Net Profit per Trip is not correlated with the other quantities, which are strongly related to each other as expected.

This means that Net Profit per Trip is a good informative metric, and since it is independent of fleet size, the number of active users, trips, and revenues that are made in a city, it must depend on some external factor (such as the average trip duration, which maybe depends on the size of the geofence, or the habits of the inhabitants of that particular city).

Fig. 2

How to use the Net Profit Per Trip

We know that Net Profit per Trip is independent of other quantities, and we can estimate it around 2.5€ per trip, according to the dataset we analyzed (graph) and the estimation previously made. So, this number can change according to possible fixed costs variations, and it is not intended as specific for the use case. 

We know that fleet and users are correlated, as well as users and trips, as shown in the figures below

Fig. 3
Fig. 4

The Net Profit per Trip is crucial to start a sharing service because it can be used as a starting point to calculate the number of vehicles needed for the service to have a certain amount of monthly gain. 

For example, suppose you want to earn 10K€ per month (Net Profit). 

We know that the number of trips needed can be evaluated starting from the Net Profit, divided by the NPT, so 10000/2.5 = 4000 trips. 

From figure 4, we know that for 4000 trips are needed about 1100 users, which according to figure 3 are obtained with about 60-70 vehicles. 

Nothing more than that. 

This is not as random as it sometimes seems. You just have to know how to read the numbers

Disclaimer ❗️

All this needs to be parameterized and scaled, considering that the analysis looks at the month with the most trips in the year, but it is the information that the analysis contains that is crucial, much more than the final numerical result. Finding a rule that can give our customers an idea of how many vehicles they need, answering the million-dollar question that no one ever answers.

About the author


Leandro Nesi

Data Scientist at 2hire

Less is more. I look for science and numbers in all my interests and emotions with the people I surround myself with.