Inside the Algorithm Pricing Your Uber Ride
A new investigation found Uber and Lyft fares can differ by half for the exact same trip. Congress is asking why. The answer fits a pattern this newsletter calls Coercive Capitalism.
In March and April, Consumer Reports asked volunteers nationwide to try something Uber and Lyft likely never expected: open the app, request the same ride at the same time from the same place, and compare what happened. The fares were often very different. Lawmakers from both parties have since sent letters asking for explanations. The companies insist it is just normal competition. This article considers what these companies know about you before setting your price.
Three Things to Know
Consumer Reports had 174 volunteers in 18 states test Uber and Lyft. They found that fares for the same rides frequently differed by about 50 percent.
The House Oversight Committee and a bipartisan group in Congress are now officially asking the companies to explain how their pricing algorithms work.
Uber and Lyft control about 95 percent of the ride-hailing market. If you do not like their answers, there are few other options.
Surge pricing told you the market had changed. Surveillance pricing tells the market that you changed, and prices you accordingly.
The forty-dollar question
Let’s begin with the numbers, since they tell the main story.
In Austin, volunteers who requested the same route within minutes saw fares from $25 to $65. That’s a 160 percent difference. In Kansas City, 55 people checked the same Lyft route at the same time and got 29 different prices. In Phoenix, 18 Uber riders saw base fares around $55 to $60, but after discounts, they paid between $41.21 and $56.96. In Atlanta, 37 Lyft riders requested the same trip at the same time. Their starting fares ranged from about $12.93 to $14.99, and after discounts, they paid anywhere from $2.28 to $14.99.
In Florida, one woman was quoted $95 for an Uber ride, while another person checking the same trip at the same time saw a fare of about $65. House Oversight Committee, in a separate review released earlier this year, cited a report showing Uber pricing the same product differently for different customers by an average of 11 percent, with one documented case showing a 221 percent gap between two riders on the same trip, one quoted $76.82 and the other $23.92.
There are also fake discounts. Consumer Reports found that almost 11 percent of advertised “discounts” on both apps were based on inflated reference prices rather than real markdowns. For example, one volunteer, Tessa, saw her fare listed as discounted from $82.08 to $65.95, with a banner saying fares were “lower than usual.” But 40 other riders checking the same route saw prices ranging from $65.93 to $65.99, with no discount banner. Her so-called discount was just the normal price made to look like a deal.
What the companies say, and what researchers found anyway
Both companies deny using your personal data for pricing. Uber says it “does not personalize prices, period,” and blames price differences on real-time market conditions, GPS accuracy, and rapidly changing demand. Lyft has made almost the same statement: “We do not engage in surveillance pricing. Period.”
Consumer Reports designed its test to rule out timing as a factor, requesting rides within minutes or even seconds of each other from the same spot. Derek Kravitz, the lead author, pointed out that Uber and Lyft collect a lot of behavioral data, like how quickly and accurately you type an address. This data could be used to guess how much you are willing to pay. Researchers have seen this before. One study in Chicago found that no two riders, even those requesting a trip milliseconds apart from the same place, were likely to get the same price. Fares were often higher for trips linked to lower-income and mostly nonwhite neighborhoods. Another analysis found Uber charged more for trips starting or ending near expensive hotels.
The architecture underneath the fare
At this point, the story is no longer only about one app. It is about the entire industry.
In July 2024, the Federal Trade Commission began a formal study of what it calls surveillance pricing. This is when companies use your personal data to set a unique price for you rather than a single price for everyone.
The Commission’s staff lists the data used: your location, demographics, browsing and purchase history, device type, battery life, and even how you move your mouse on a screen.
In March, the House Oversight Committee sent letters to Uber, Lyft, Booking, Expedia, and Instacart. They warned that this kind of data can be used to find a consumer’s “pain point,” which is the highest price someone will pay before deciding not to buy.
This request for information is separate from the FTC’s 6(b) study, which has subpoenaed consultancies and tech vendors that create pricing tools for other companies.
This investigation is not focused on rideshare, and there is no evidence that it involves Uber or Lyft. But it does show that this kind of pricing system exists at scale, built by companies most riders do not know about. These systems are built to find the highest price a person will pay, not the price the market would set for everyone.
This is the big change. Surge pricing reacts to things like more riders, fewer drivers, or bad weather. Surveillance pricing, however, is based on who you are, what device you use, how you act in the app, and how easy it is for you to switch to a competitor.
This also explains why most people do not notice. A study found that only about one in six riders checks both apps before booking, and the average price difference between Uber and Lyft for the same trip is about 14 percent. Sticking with one app is a habit that pricing algorithms can exploit to charge you more.
Running it through the test
This newsletter uses a four-part test to distinguish between normal business pricing and what it calls Coercive Capitalism. First, the deal must look voluntary. Second, the customer must get real value, not just the appearance of value. Third, the data you give up in the process must be used against you. Fourth, the cost of leaving must be high enough to be more than a small hassle.
Uber and Lyft meet the first two points. No one makes you use the app, and when it works, you get real value by getting from one place to another.
The fourth point is the hardest to argue with. Two companies control about 95 percent of the market, leaving almost no other choices. The way prices are set is a secret, so riders cannot check or predict them. When you really need a ride, like when it is raining, you are late for a flight, or you are leaving a hospital, you are least likely to compare prices. People who rely most on these services, such as those without a car or with limited flexibility, are least able to check both apps. This is how the K-shaped economy shows up in a taxi fare. Avoiding a $40 extra charge is not the same as being protected from the system that causes it.
The market is not broken here. It is working as it was designed to, and that design is the real story.
What You Can Do
THIS WEEK
Before your next ride, open both apps and compare the price for the same trip before you book.
Take a screenshot of the fare you see. If the price changes after you close and reopen the app, save that screenshot as well. Check it against the price other people near you are seeing for the same trip. A discount off an inflated number is not a discount.
THIS MONTH
Send a comment to the FTC’s open surveillance pricing docket, or write to your member of Congress and mention the Consumer Reports findings. Both the House Oversight Committee and the Monopoly Busters Caucus are collecting public input, and your comments become part of the official record.
Keep track of your ride receipts for a month, noting the pickup point, time, and final price. One trip will not tell you much, but a month of your own data is the same method Consumer Reports used, just on a personal level.
Stop using just one app out of habit. The data shows that this loyalty is already factored into prices. Comparing both apps each time takes only thirty seconds and makes you less predictable to the algorithm.