Fair Pricing for Reliable Rides
Every day, people do millions of rides to move around in their city, but there are never two completely identical trips. Likewise, the same user can have a different ride experience every time they hail a ride.
At Taxify, we want to deliver a great ride-hailing experience for everyone — both our passengers and drivers. Although there are tens of aspects that build up into a travel experience, we know that most users care about two things: the availability and pricing of their trip.
Recall the last time you wanted to get to the airport early in the morning. Or the time when you were looking for a car to leave a busy event venue. In such occasions, you’re expecting to find a vehicle that arrives fast, and take a trip at a reasonable cost.
Having a limited number of drivers providing their services and a potentially unlimited demand for rides, it is a worthwhile task for our team of data scientists to figure out how to apply the laws of economics and ensure that everyone has the best possible ride experience.
Let’s take a closer look at the components that supply the bones of fair pricing and reliable ride-hailing experience.
Dynamic pricing
If you’re like most people, you value the convenience of a fast-arriving ride. In fact, we see long ETAs (Estimated Time of Arrival) significantly increasing the rate at which passengers cancel ride requests.
The underlying question is how to set a fair price for each ride while ensuring that a vehicle is always available when you need one. In high-demand areas and on high-demand hours, the answer lies in dynamic pricing.
If many people in a specific area want to get a ride, raising the prices will motivate more drivers to move towards the hotspot. Some passengers may also decide that getting a ride at a heightened price is not worth the extra cost and they would rather wait for a bit longer. As a result, passengers with a strong need for a lift can rely on their app to get a ride even if the demand in their surroundings is currently high.
One potential issue with dynamic pricing is that it is reactive by its nature. Prices will surge and drop only after the changes in local demand, lagging behind the real-time need. This may cause a situation when for instance an area in the city lights up with surge pricing, but by the time drivers arrive on the spot, the need for rides has already dissolved.
Here is an example of how the demand in the city may change over 15 minutes. Yellow colour represents higher demand, green means less demand.
On a closer look, you can see demand for rides increase in some areas while decreasing in others.
To solve the discrepancy, we’re developing intelligent demand prediction — both short- and long-term. Based on sufficient historical data, it is possible to predict when and where in a city there’s a heightened need for rides. Confidence in the prediction increases as we augment our in-app data with the outside world’s datasets. For example, weather forecasts and the schedule for large-scale events help to predict a need for more drivers in specific city areas.
This way, we can apply dynamic pricing ahead in time to ensure that a ride is always available, even during busy hours and across active areas.
Upfront pricing
We believe that both passengers and drivers should have full transparency, knowing how much they’re about to spend or earn. Knowing how much a trip is going to cost gives you a peace of mind.
The tricky part about putting a price tag on a route is that the exact cost is known only after the ride has ended. That’s when upfront pricing enters the game, with the goal of providing the exact route price already while ordering a ride.
Based on past route and pricing data, we’re training machine learning models to predict the exact cost for similar rides. Add on top a layer of predicting GPS errors, driver behaviour, and potential detours; and you can predict route costs at a fairly reliable accuracy.
Route-based pricing
To ensure fair pricing for everyone, both passengers and drivers need to be taken into account.
Every time you take a ride towards the city centre, your driver has a higher chance of finding the next passenger. Correspondingly, when a driver takes you further from high-demand areas, they may need to drive a longer way back to find their next customer.
Route-based pricing helps to ensure the supply-demand balance across vast urban areas. For example, as a passenger, you may get a slight discount on your trips from low-demand to high-demand areas. As great ride-hailing experience depends on both pricing and ETA, you may also be charged less when your pickup takes more time to arrive.
Bringing change in increments
Every single ride’s price is a sum of several components: dynamic pricing, demand prediction, upfront pricing, and the route in question. At Taxify’s Data Science team, we’re working on improving every aspect of the price and route calculation.
And to be fair, this list of aspects contributing to ride pricing is by no means finite. Pricing is an extremely complex topic with many interdependent elements, and we’re constantly exploring new directions to provide fair pricing and reliable ride-hailing experience for everyone.
If you feel like being part of a fast-moving Data Science team building a technology platform for millions of people around the world, check out our job openings and join the ride.
About the Author
Mikhail Iljin is one of the first members of Taxify’s Data Science team. His main responsibility is developing real-time data-driven pricing for passengers and drivers.
Previously, Mikhail has worked as a Senior Software Engineer and Data Scientist, working and consulting on products in Swedbank, Estonian Digital Certification Centre, as well as creating data analysis and warehousing solutions for Telia, Tele2 and EU-funded urban mobility research project in collaboration with Tartu University.
