You’re in Manhattan’s Times Square, run­ning late for dinner at Le Cirque, on East 58th St. You open the Uber app on your smart­phone, hoping a car from the now ubiq­ui­tous ride-​​sharing ser­vice is nearby, only to dis­cover that you’ll have to pay 1.5 times the base rate for the ride.

New research, led by Christo Wilson, assis­tant pro­fessor in the Col­lege of Com­puter and Infor­ma­tion Sci­ence, unlocks details behind the algo­rithm that drives this surge pricing. Based on data from Man­hattan, Wilson’s team also has a fix:

Wait five min­utes, or walk a few short blocks, and the surge noti­fi­ca­tion may disappear.

Under the Uber hood

On Thursday, at the 2015 Internet Mea­sure­ment Con­fer­ence, in Tokyo, Wilson revealed that and more of Uber’s secrets.

Wilson and his co-​​authors, Le Chen, PhD’16, and asso­ciate pro­fessor Alan Mis­love, had long been trou­bled by Uber’s “lack of trans­parency.” Other sharing mar­ket­places, such as Ebay and AirBnB, openly dis­play their prod­ucts and prices online, enabling cus­tomers to make informed choices. Uber, on the other hand, oper­ates in the dark: It releases no num­bers about how many people are requesting cars or how many dri­vers are avail­able, and prices vary wildly, based on time and place of cus­tomers’ requests.

You have to trust that their system is doing what they say it’s doing,” says Wilson, whose schol­ar­ship focuses on auditing algo­rithms. “But that’s the ques­tion: Is it doing what they claim?”

Well, yes…and no, the researchers found.

Research by Northeastern's Christo Wilson reveals Uber pricing disparity in New York's Times Square

To bal­ance supply and demand during periods of high usage, or “surges,” Wilson says, Uber uses “an opaque ‘surge pricing’ algo­rithm” that changes fares every five min­utes. And it divides the cities it ser­vices into “dis­crete ‘surge areas.’” The con­flu­ence of ric­o­cheting prices and dis­cretely defined areas leads, the team found, to an unusual—and unfair—scenario: “corner cases,” says Wilson, “where you can walk across the street and all of a sudden the price changes.”

Times Square, their data show, is one of those corner cases. “For example,” they write in the paper, “20 per­cent of the time in Times Square, cus­tomers can save 50 per­cent or more by being in an adja­cent surge area.”

What exactly are surge areas? They are man­u­ally demar­cated sec­tions of a par­tic­ular city, each with its own inde­pen­dent price based on inten­sity of demand at a par­tic­ular point in time. The maps of the surge areas look like funky jigsaw puz­zles: uneven pieces with mostly sharp edges locked together. Boston, for instance, has nine surge areas. Man­hattan: 16. London: 19. Still, says Wilson, how Uber divided up the cities is not clear.

Uber surge regions: Boston

Uber divides the cities it ser­vices into “dis­crete ‘surge areas,’” according to research led by Northeastern’s Christo Wilson. Boston, for instance, has nine such regions. Graphic by Erica Lewy/​Northeastern University.

Becoming Uber

To con­duct the research, Wilson and his col­leagues did more than look under Uber’s hood; they essen­tially crawled inside its com­puter systems.

Using servers in a closet on the North­eastern campus, they pro­grammed and ran Uber apps “pre­tending” to be people at 43 dif­ferent GPS loca­tions throughout San Fran­cisco and Man­hattan over a four-​​week period. The researchers chose San Fran­cisco and Man­hattan for sev­eral rea­sons, including their having, respec­tively, the second and third largest number of Uber dri­vers in the U.S. and large dif­fer­ences in access to public transportation.

If you’re a cus­tomer, it can pay to wait or walk.
— Christo Wilson, North­eastern Uni­ver­sity assis­tant professor

The data they col­lected included the surge price—that is, the number by which the base price was mul­ti­plied during surges—and esti­mated wait time for each “ride” as well as the loca­tion of the “request.” Crunching the data, they tracked supply and demand, how those dynamics changed over time and dis­tance, and the way surge prices varied by location.

We did a lot of cor­re­la­tion analysis looking at how many cars were get­ting booked over time and how many cars were avail­able, and you do see high cor­re­la­tion between supply and demand and the surge,” says Wilson. “So the system is def­i­nitely responding to supply and demand changes.”

Christo Wilson, assistant professor in the College of Computer and Information Science, led the first in-depth investigation of Uber, the ride-sharing service. Photo by Mariah Tauger/Northeastern University.

Christo Wilson, assis­tant pro­fessor in the Col­lege of Com­puter and Infor­ma­tion Sci­ence, led the first in-​​depth inves­ti­ga­tion of Uber, the ride-​​sharing ser­vice. Photo by Mariah Tauger/​Northeastern University.

Still, there is room for manip­u­la­tion, on the part of both dri­vers and customers.

The data wasn’t con­clu­sive regarding dri­vers. But Uber driver forums, Wilson reports, con­tained con­ver­sa­tions about “collusion”—drivers in a spe­cific area uniting to go offline to arti­fi­cially reduce supply and thus lead to a price surge. “The dri­vers talk about this, but we don’t have any evi­dence that this actu­ally works,” says Wilson.

For cus­tomers, how­ever, the find­ings were explicit. “If you’re a cus­tomer, it can pay to wait or walk,” says Wilson. He notes, how­ever, that you can’t know where to walk to or how far if you don’t have a detailed surge-​​area map in hand—and Uber is unlikely to ever pro­vide one.

How­ever, Wilson and his team will. “Sit­ting on my com­puter is all the research we’re doing in this vein,” he says. “We have a web­site for it, and even­tu­ally we’ll have a page about the paper that’s acces­sible to the public. We are devel­oping surge maps, and will put all of them there there, too.”