“You don’t own your car; it owns you!” – It’s an idea being voiced more and more often these days. Many of us are aware of the problems associated with owning a car – the cost, the hassle, the environmental damage, et cetera. In fact, if we add up the costs, including road tax, repairs, fuel, parking, and depreciation, we find that the average car owner in the Netherlands pays about 520 euro per month for their car.
But we’re not ready to get rid of our cars just yet. One of the main reasons is because we like having a car always available, in case we need it right now. We want a guarantee for mobility.
The question is, “Can mobility ever truly be ‘on-demand’ like music or TV, á la Spotify or Netflix?” We think it can.
But in order to provide reliable “mobility as a service” for people using cars, we need to essentially see the future. We need to be able to predict where and when people will need cars, so that we can relocate cars to the right location (that’s why we want to develop self-driving cars as well).
The Crystal Ball
To “see the future”, we teamed up with Pipple, a small data science consultancy company in Eindhoven. Their team of creative econometricians help clients to solve complex data problems. In more technical terms, what Pipple helped us generate was a forecasting model, which predicts where and when our clients need cars.
Forecasts are always built on historical data, to some extent. But what to do when the data isn’t there? After all, in 2016 our company was young, and we hadn’t even put the first cars into service yet. “For Pipple, the main challenge was to overcome the lack of large amounts of car user data, which we needed in order to develop the forecasting model,” relates Wouter Leenen, data scientist at Pipple,
“To tackle the challenge, we developed a subsystem to keep track of all our data,” explains Joep Sloot, our COO in charge of heading up the project from our end. But this was easier said than done. “In the beginning, the main challenge was figuring out which data was needed in order to forecast demand. We also needed to know which internal and external factors influenced this demand.”
“Amber mainly focused on gaining data for Pipple and creating the platform,” adds Erlijn Linskens, another Pipple data scientist. “The main task for us was analyzing data and afterwards developing the forecasting model. In the final phase, Amber was responsible for the interplay between the data, the system and the forecasting model.”
In the end, Pipple developed two algorithms for us: One for short-term use, and one for use in the future. The short-term algorithm we currently use in our B2B car sharing platform. This is the algorithm that makes sure that employees at clients such as ABN AMRO will always have a car available when they need to go to a client appointment. Since we eventually plan to incorporate private users as well, we also needed a more sophisticated predictive algorithm “on the back burner”, so to speak (read about our entire vision here).
A Tale of Two Startups
It wasn’t all numbers and figures during the collaboration, however. “We discovered that we actually had a lot in common,” says Wouter. “We were both startups, with similar working cultures. We had fun together!”
Our collaboration with Pipple enabled us at Amber to take our product “to the next level”, for lack of a less-cliché phrase, and allowed Pipple to gain valuable experience in the mobility and transportation industry.
Oh, and did we mention that Pipple is now using Amber? It often doesn’t make sense for a small company to own a car or be tied down to expensive leasing plans, and in any case, most of Pipple’s employees come to work by bike (many don’t even own a car), so Amber is perfect for them to use to go to appointments with clients, or even to run errands during the day.
Here’s to a lasting partnership!