DiDi is a relatively new entrant to the Australian market, with 2019 being their big coming-out party. The ride-sharing company from China is making big moves for market share and are internationally are amazingly successful, so pay attention.
When you interface with DiDi, as a driver or passenger, you think about vehicles, but at its core, DiDi is a data platform, that happens to transport people.
The company racks up a staggering 10 billion trips per year and at that scale, no human can make sense of it, which is exactly why DiDi is leveraging Artificial Intelligence.
DiDi built an AI engine that processes more than 4,875TB of data per day, while receiving over 40 billion daily route planning requests. These routes are the journeys drivers follow to deliver passengers to their destination in the fastest possible path.
With roads and highways constantly changing, the challenge of accommodating those changes is not small when dealing with this scale of operations. A staggering more than 106TB of new location and route data is added to the system, each and every day.
At DiDi AI Labs, engineers and data scientists collaborate to improve the accuracy and efficiency of DiDi’s big data algorithm which powers the user experience. The Lab focuses on machine learning, natural language processing and computer vision and works with a range of partners including Stanford University, University of Michigan and Hong Kong University of Science and Technology.
In Australia, DiDi is available in Melbourne, Geelong, Newcastle, Brisbane, Gold Coast, Sunshine Coast and Perth. It’s perhaps the global story that’s the most impressive, with the company’s 11,000+ employees servicing some of the largest populations on the planet, including China, Mexica, Japan, Brazil, Colombia, Hong Kong, Chile and Costa Rica. In total, they have more than 550 million users across over 400 cities.
Supply and Demand Forecasting
DiDi’s AI engine (also known as DiDi Brain) is used to develop big data strategies for efficiency and safety architecture through machine learning, and cloud computing technologies.
Algorithms to predict passenger demand and driver supply in different areas at any given time, enables the service to provide the best customer experience possible and lower overall costs per ride on the platform. These predictions are based on trip request volume and driver location.
This demonstrates DiDi’s understand of one of the most important metrics for users, that’s time-to-pickup and minimising that time, can be the difference between winning or losing a customer.
Smart Route Planning
Leveraging a large volume of data means DiDi can design better route planning algorithms, which better simulate the decision-making processes of experienced drivers and bring users the most efficient mobility options while enhancing overall traffic effectiveness.
Given how much of an issue congestion is in our major cities, routing is absolutely critical to the service working correctly and ensuring the passenger has a great time, every time.
Pickup probability models leverage historical data to ensure drivers are in the right areas at the right time of the day, before trips are even booked on the platform. This system increases the match success rate between drivers and passengers to enhance overall transportation efficiency and improve the user experience.
Sharing a car with others, reduces the cost for each rider, but presents an additional challenge to the routing algorithm. Should you plan a path that increases journey time, but has the chance to reduce the cost?
Algorithms efficiently match carpool orders on similar routes in real-time so that users can share transportation resources and reduce the price.
DiDi says ever more accurate pool models are on the way and will further enhance rideshare experiences and efficiency of their services. Based on July 2019 data, DiDi finished 1 billion car-pool trips per year, and help consumers achieve 25% savings while enhancing driver income.
When passengers choose “Express Pool”, DiDi’s system will calculate 3 to 5 possible pickup points based on multiple metrics, such as walking distance, time needed, and traffic conditions, then recommends the most suitable one.
After a driver accepts the order, the driver and passenger have to arrive at the station within a certain time, which is determined by the station location and real-time traffic conditions. This feature further optimizes route planning and improves transportation efficiency.
Based on real-time travel data, time estimation algorithms are overcoming the shortcomings of traditional computing methods. This provides a more accurate time of arrival estimates, ensuring you can communicate more precisely communicate to others, like attendees in a meeting, when you’ll make it to the office.
AI-Powered Driver Voice Assistant
DiDi’s Australian scientists and working with DiDi Japan to develop a hands-free app interface that enables drivers to complete passenger pick-up and route navigation using only voice commands. This feature is currently in the testing and development phase in Melbourne.
As you know, on other ride-sharing platforms, part of the pickup process is the driver calling you to confirm they have arrived, or a precise meeting point, which side of the road you’re on.
This call is often done by the driver reaching the destination, pulling over, then making a call to the customer. This is inefficient and that inefficiency can easily be overcome by removing the need to touch or focus on their mobile device.
AI-Assisted Arbitration & Dispute Resolution
If you run a ride-sharing service, you’d hope 100% of the customers were happy, but with the size of the DiDi customer base, even 1% of 1% is still a big number.
DiDi develops have created an AI-powered system to settle driver-passenger disputes in a fairer way during the feedback review process.
For instance, when an order is cancelled either by the driver or the rider, DiDi’s order cancellation and responsibility division system can effectively decide which party is responsible by considering multifaceted factors such as the order time, estimated time of pick-up, real-time location of the vehicle, road conditions, etc.
It takes just 10 milliseconds for the system to make a decision, and the accuracy rate is 93%, which leads the industry.
If the system decides against a driver, it could negatively affect the driver’s ratings. If the system decides a driver is not responsible for a cancellation, the driver’s turnover rate and ratings won’t be impacted. If a driver or rider has doubt about a decision, he/she can file an appeal and demand a review, ensuring even greater accuracy and fairness in the system.
As you can see DiDi is a service that leverages data heavily to optimise business operations and improve the customer experience. This strategy has helped them achieve some seriously impressive growth since the company began back in June 2012.
DiDi is also the official rideshare partner of Star Wars: The Rise of SkyWalker in Australia. You can Ride Smart by summoning a DiDi and get $20 off rides using the code STARWARS (only for new users).