But at the company’s end, running an effective service involves a number of complex issues. Among them: How can companies encourage people to share rides with strangers, which lowers costs and carbon emissions, rather than having the car to themselves? And, when drivers are expecting new requests, how should they decide whether to stay put or head to another neighborhood?
Two studies from the Kellogg School have addressed these questions. The first explored how best to address one of the reasons customers often choose to travel alone: ​​detours to pick up a colleague. The researchers developed a mathematical model showing that it is possible for companies to significantly reduce detours on shared routes and improve efficiency, and that a key element of this is getting more people to use this type of service.
A second group of researchers looked at how companies could more efficiently route drivers waiting for ride requests. Using an optimized algorithm to make these decisions, they found, could greatly increase the chances of winning customers quickly.
“This is something that needs to be planned and designed carefully,” he says Anton Bravermanassistant professor of operations at Kellogg, who worked on routing research.
No free ride
The first post focused on shared rides because sharing a car with another customer has many advantages over solo rides. They are cheaper for passengers and cause less traffic congestion and produce less carbon emissions. (Common routes are currently suspended in many locations due to the risk of transmission of COVID-19.)
Even before the pandemic, however, many people chose to drive themselves because they didn’t like the driver going out of their way to pick up or drop off another person. Car shipping companies generally try to do this race passengers who want to go in the same direction. But sometimes, even if the second passenger is very close, “you have to turn left four times just to pick someone up because you’re going around the block, which can be very frustrating,” he says. Sebastian Martinassistant professor of interventions at Kellogg, who conducted the research while a postdoctoral fellow at Lyft.
Transport companies have taken a number of measures to minimize detours. For example, Uber offers a ride-share option called Express Pool and Lyft offers Shared savings, which allows customers to pay an even lower price if they are willing to walk a few minutes to a pick-up point that is convenient for the driver. The customer also disembarks near, but not necessarily directly to, their destination in order to stay on a more direct route for other passengers.
But Martin wanted to understand the relationship between bypasses and value more rigorously. The price, in this case, meant the efficiency gained from carrying two passengers on a shared route instead of carrying each person in a separate car, measured as the reduction in total driving time.
Martin conducted the research with Ilan Lobel at New York University, who was a visiting researcher at Lyft Marketplace Labs at the time. To investigate this, they created a mathematical model of the process of matching passengers to shared routes. The researchers confirmed that bypass and value have a negative relationship. That is, when the bypass is low, the value is high.
“If we match people well, with high value, they’ll usually also have low churn,” says Martin. “You can have the best of both worlds.”
Two-tier system
So how should a company decide whether to match two riders? In an extreme case, its algorithm can allow zero detours, bringing passengers together only if those people are traveling along the exact same route. Or the company may prioritize value maximization, regardless of how much bypassing it involves.
Martin and Lobel tested various permutations using their mathematical model to predict the effects of prioritizing high-value or low-value detours and how these would change as the number of route requests increased. The researchers conducted the analysis on several types of hypothetical city structures, including a grid, a tree (similar to a city center with paths branching out), and a simplified freeway. They also conducted a separate simulation of the city structure of Manhattan, incorporating data on approximately 847,000 actual trips using transit services in that area.
The researchers found that even if the company’s route-matching algorithm allows for a small detour, the value can be very high. But the key factor, the team found, was the number of people using the driving system. If the number of customers doubled, the bounces dropped dramatically while delivering the same value.
The results led them to propose a new system: offering two types of shared-route services in order to attract as many people as possible. The first, more expensive option will guarantee very little detour in order to attract customers who hate going out of their way. The second, cheaper option would allow more detours and attract people who want to save money and don’t mind spending more time in the car to do so.
“The name of the game is to get people to use these shared services,” says Martin.
Driving in the Void
Braverman and his colleagues looked for another opportunity to improve efficiency. When a car drops off a passenger and there are no other immediate requests in the area, the driver must make a choice. Do they hang around and wait for a nearby request to appear? Or are they headed to another neighborhood in the hope that requests there might be more frequent?
If this “balancing” of cars isn’t done right, “you just have drivers driving around on empty for no good reason and wasting money on gas,” Braverman says.
Right now, these decisions seem to be somewhat ad hoc. Uber incentivizes drivers using surge pricing. If a particular neighborhood has a lot of requests, the company increases the pay rate in that area. But beyond that, it seems drivers “just do whatever they want,” says Braverman, who worked with JG Dai at Cornell University and Lei Ying and Xin Liuboth at the University of Michigan, about the study.
The researchers investigated how companies could develop a better algorithm to guide drivers. They created a mathematical model and wrote software to simulate a ride service, taking into account factors such as the frequency of ride requests in different areas and traffic congestion. The goal was to maximize availability—that is, the probability that when a customer requested a ride, a driver was around to pick them up.
The computer program solved the optimization problem and directed the drivers to either stay or go somewhere else after dropping off their last passenger. (Although there’s no simple way to describe how the algorithm makes these decisions, the program is simple enough that a company could easily implement its own version, Braverman says.)
Smarter routing
Next, the team wanted to know how much their algorithm improved the system’s efficiency. To calibrate their model with real-world data, the researchers obtained details of about three weeks of rides with Didi Chuxing (a service similar to Uber) in an unnamed city in China. They then ran computer simulations of a typical day.
The simulations allowed them to compare the performance of their algorithm to basic heuristics that drivers could use without explicit instructions from the transportation company. For example, perhaps drivers are trying to predict where they will most quickly find another customer, taking into account factors such as how long it takes to drive to that particular neighborhood and the likely frequency of requests. If they followed this rule of thumb, about 30 percent of customers’ transfer requests would go unfulfilled, the team found. But if drivers followed the algorithm developed by Braverman’s team, that number would drop to 20 percent.
However, the team’s algorithm assumed that traffic patterns and request rates remained the same throughout the day. In a more realistic simulation where the rate of route requests varied at different times, the algorithm did not perform as well during rush hour. For example, from 7 to 8 p.m., 33 percent of customer requests were not fulfilled.
So they devised another heuristic called “look-ahead policy,” which predicted future traffic and requests based on past data and routed cars accordingly. With this policy, unfulfilled requests during peak hours were reduced to 14%.
The model still needs more work to be realistic enough for real-world use, Braverman says. However, the study suggests that properly balancing the cars is an important part of efficient operation.
“You should reposition yourself smartly,” he says. “Because if you’re not, you’re leaving a lot on the table.”
