Comment on Consumer Reports Study

Consumer Reports based its findings on a flawed methodology: what they describe as the “same” trips are actually different trips that were priced differently due to changing real-time marketplace conditions, not personalization. 

Different trips requested at approximately the same time will have prices that are approximately the same—which is exactly what their data shows.

Consumer Reports’ methodology was built to produce the findings they sought, regardless of how Uber’s pricing actually works.

###

Recently, Consumer Reports published a report based on a small-scale experiment attempting to analyze Uber’s trip pricing, promotions, and platform take rates. While we welcome outside groups studying rideshare platforms, the publication’s findings are based on a flawed methodology, a non-representative sample, and a fundamental misunderstanding of how an open and dynamic rideshare marketplace works.

To be clear, Uber does not engage in surveillance pricing, we do not personalize prices to individuals, and we do not use protected characteristics, phone battery levels, phone models, or other device information to set prices. User-specific behavioral attributes or customer segments also do not factor into rider prices.  

We’ve read the report closely, and—even without access to the full underlying data, which Consumer Reports declined to share—we are confident that their methodology does not support their findings:

• These are not the same trips: Consumer Reports treats trips with the same pickup and drop-off points as identical. They are not. In a real-time marketplace, a trip is defined not only by where it starts and ends, but also by when it is requested and what marketplace conditions exist at that exact moment. Rider demand, driver availability, traffic, routing, and estimated trip length can all change within seconds. The report’s analysis therefore compares different ride requests made at different times under different conditions, not the same trips. This core flaw underlies every finding in their report and shapes the headline numbers that follow. 
• Uber’s promotional discounts are not fictitious: The methodology behind their claims about “fictitious discounts” relies on another flawed assumption. Consumer Reports arbitrarily decided that a trip can only be genuinely discounted if it is cheaper than the median price for a given route. But a route-level median is not the correct reference price for a specific trip request. Trips with the same origin and destination can legitimately have different prices because real-time marketplace conditions change. This is not evidence of fictitious discounting—it’s an expected feature of a dynamic marketplace where non-personalized pricing inputs fluctuate continuously in real time.
• Uber’s U.S. take rate is around 20%: The report’s findings on take rate were distorted by a flawed experimental design. Volunteer riders and drivers were placed in the same room—but driver earnings depend in part on how far the driver must travel to reach the rider. By minimizing pickup distance, the experiment created an artificial scenario that isn’t representative of reality. Adjusting for insurance costs, which we voluntarily carry on behalf of drivers, our average revenue per trip is around 20% in the U.S. and has remained at that level for years.

We want to set the record straight by breaking down where this study falls short.¹

Consumer Reports’ False and Misleading Methodology

The researchers claim that Uber routinely charges different customers different prices for the “same” ride. But their findings are not supported by their methodology. 

The central flaw of the report is that it treats trips with the same pick-up and drop-off points as the “same trip.” In reality, those are different trips. In an open, dynamic marketplace, a trip is defined not only by where it starts and ends, but when it was requested and the marketplace conditions at that moment. Uber’s pricing is marketplace-driven. Fares update automatically in real time based on objective factors like rider demand, the estimated time and distance of the trip, real-time traffic conditions, and the immediate availability of nearby drivers.

Consumer Reports’ data collection procedure does not precisely control for time. Volunteers were asked to request fares for similar trips in a poorly controlled environment, where it was impossible to ensure that those requests would happen at exactly the same time or under the same marketplace conditions. 

The timestamp data that Consumer Reports used comes from when screenshots were taken, rather than from when a fare request was actually sent to Uber’s systems. The timestamp on a screenshot is only accurate to a minute, yet Uber’s marketplace can update within seconds. Uber also temporarily ‘holds’ the fare shown to each person to make sure the price isn’t changing while they are considering their ride options. That means two people could request fares at different moments, receive different prices based on real-time conditions, and then submit screenshots showing the same minute—even though the fare requests happened at different times.

The researchers told us that “to attempt to control for time, we had our volunteers open, review, and screenshot their offer screens at roughly the same time.” Volunteers screenshotting offers at “roughly” the same time will see prices that are roughly the same—which is exactly what their data showed.

Second, Consumer Reports drastically overstates the significance of their own (flawed) findings. The price differences in the dataset, caused by the factors above, are generally small. Even in the partial data they shared with us, over 80% of ‘original’ fares are within 5% of the median. The group’s finding that “14 of the 15 routes tested across the country had at least two price clusters separated by at least 5%” describes the same pattern: prices clustered close together for similar trips, with small differences that likely account for real-time variables. Given the operational complexity of a real-time marketplace, that pattern shows broadly consistent pricing, not personalized or differential pricing.

Finally, the experiment itself likely affected the results. The tests performed involved participants requesting fares for similar rides within compressed time windows. In a real-time marketplace, concentrated bursts of simultaneous fare requests can themselves affect prices. To our system, the experiment could have looked like a spike in demand, and a group of participants who start requesting fares just a few moments later may see different prices for that reason alone. That is exactly how dynamic pricing works—it responds in real time to signals of increased demand. 

The report also relies on the average difference between the maximum and minimum trip price they identified within similar routes. This measure is highly sensitive to outliers caused by limitations of the experiment’s setup, as well as possible data processing errors, and cannot support any findings about pricing variability.

How Pricing Works at Uber

Price variability for seemingly identical trips is driven by operational factors, not by personal characteristics. These factors can include: 

• Timing Sensitivity: Prices update second-by-second. Even minor variations in app load times or the exact moment a request is initiated can result in different price inputs.
• GPS and Routing Precision: Upfront pricing depends on GPS coordinates to estimate pickup and drop-off points. Due to inherent margins of error in smartphone GPS technology, two devices located side-by-side can report different pickup coordinates. These slight location variances shift the estimated routing and pickup time, which in turn adjust the final upfront price.
• Real-Time Marketplace Dynamics: Pricing includes an assessment of the driver’s effort needed to reach a pick-up point. Predictions regarding nearby driver availability change dynamically; if our system anticipates a driver needs more time to reach a specific location, the upfront price may adjust to reflect that.

The key point is that differences in price for seemingly identical trips are driven by operational and external factors like timing, GPS estimates, and nearby driver availability—not the use of riders’ personal characteristics, such as demographics, device type, or battery level, or any form of “surveillance pricing.” 

How Promotions Work at Uber

A promotional discount always reduces the fare for that trip request. When a promotion applies, the rider sees it before confirming the ride: their original fare is explicitly crossed out and replaced with the discounted price. These discounts are transparent, shown up front, and reduce the fare the rider would have otherwise paid for that trip.

The report’s claim that around 10% of all discounts advertised on Uber are based on “fictitious or false reference prices” is incorrect. That finding rests on an unscientific assumption: that a median price calculated across a set of broadly similar trips can be treated as the “correct” reference price for any individual trip. This is a self-fulfilling prophecy, with conclusion baked into the methodology.

To reach their finding, the authors grouped similar but non-identical trips into “routes,” then calculated a median price for each route, using only non-discounted fares. They then compared the discounted fares against the median. Yet, as discussed above, Uber’s pricing adjusts quickly based on objective marketplace factors like demand, supply, and traffic—so trips that appear similar within a route grouping can legitimately have different prices.

That is the core flaw. A route-level median is not the “original” or “real” price for a specific trip request. It is a statistical midpoint across a set of non-identical observations. Treating it as the “real” reference price for every trip in the group ignores the marketplace conditions that may have affected each individual fare.

By their logic, this methodology would always conclude that Uber engages in so-called ‘false reference pricing’ anytime fares for similar trips show a certain degree of normal variability—even if the request times were different, or the trips in question were routed in different ways.

Under the report’s approach, if a rider received a discount, but the discounted fare was still higher than the route’s median non-discounted fare, that was evidence of “false reference pricing.” But that does not follow. It only shows that the original price for that particular trip was above the median for a broader group of similar trips. In a dynamic marketplace, that is expected.

Say, for example, the median price of several different trips on the same route was $10. If a rider booked a trip on that route for $12, but received a $1 promotion for a total discounted fare of $11, Consumer Reports would consider that a “false promotion”. But it’s not false—the price of that particular trip was $12, even if other trips on that route may be more or less expensive because of marketplace factors.

The researchers didn’t observe or control for those marketplace factors. That effectively guarantees that some legitimate discounts will be mislabeled as “false” whenever normal price variation exists. This isn’t “fictitious pricing” at all; it is an expected statistical property of the underlying pricing process driven by marketplace inputs that the authors chose not to recognize, or simply ignored.

The researchers also appear to have conflated certain product experiences with promotions. One striking example is their claim that “Fares lower than usual” banners in the Uber app are part of a promotional scheme. This banner is not a promotion and does not mean a promotional discount has been applied. It’s shown when base prices are lower than typical due to marketplace conditions—such as less rider demand than usual. It is intended to provide riders with historical pricing context, not to advertise a discount. 

Uber also sometimes shows a “Fares higher than usual” banner, which Consumer Reports seemingly did not mention despite it appearing in their data. Both banners are intended to give riders more transparency about typical prices. This confusion shows a basic misunderstanding of the data the researchers collected.

How Variable Take Rate Actually Works

Adjusting for insurance costs, Uber’s average revenue per trip—or ‘take rate’—is around 20% in the U.S. and has remained at that level for years. The report’s findings regarding Uber’s ‘take rate’ is wrong in two distinct ways. 

First, the researchers include the cost of insurance in calculating what Uber keeps, even though states mandate specific insurance coverage for rideshare activity, which is ultimately paid out as a cost of revenue. Uber maintains commercial rideshare insurance on every trip in order to protect riders, drivers, and third parties on the road. These insurance costs vary by state, but increasingly represent a large share of every fare. As of December 2025, insurance costs made up 31% of an average rider fare in California, 14% in Colorado, 22% in Georgia, 17% in Florida, 26% in Louisiana, 27% in New York state (excluding NYC), and 15% in Texas, to name a few. 

In fact, Uber’s U.S. mobility insurance costs have increased by more than 50% per trip over the past few years as of the end of Q1 2025. Categorizing that expense as part of Uber’s take rate creates an artificially inflated number.

Second, the experiment produced biased data because of how it was designed. Consumer Reports invited a select pool of volunteer riders and drivers who were placed within a few feet of each other, and then had the riders request trips until they got matched with a driver in their test group. The ‘take rate’ was then recorded.

To understand why this matters, consider how driver-side fares work. On Uber, the rider’s upfront price is locked in before they hit “request”. A factor in the driver’s fare, however, is how far they have to travel to reach the rider. A driver closer to the pickup point earns less on that component than a driver farther away because they are traveling a shorter distance.

That means Uber’s take rate on any given trip depends partly on how far the matched driver is from the rider. When a nearby driver takes the trip, the driver’s pick-up payment is small and the amount Uber keeps is higher. When a more distant driver takes the trip, the driver’s pick-up payment is larger and Uber’s take rate is typically lower. Both are normal outcomes of how the system works, and as we have previously written, this is what allows us to show upfront fares to riders and drivers.

By placing riders and drivers within a few feet of each other and matching them, the field test done by Consumer Reports only captured trips with near-zero pickup distance—a setup that isn’t representative of how Uber actually matches riders and drivers in the real world—and then averaged those take rates as if they were typical. They are not.

Concluding Remarks

Consumer Reports’ study does not support the claims it makes. Its analysis treats similar trips as identical, uses a route-level median as if it were a trip-specific reference price, and relies on an artificial take-rate experiment that does not reflect normal marketplace conditions. The methodology was built to produce the findings they sought, regardless of how Uber’s pricing actually works. 

Riders, drivers, and policymakers deserve a rigorous and open conversation about how a rideshare marketplace operates. That requires accurate data and a methodology that can be tested. Uber is committed to building a marketplace that is transparent, fair, and works for the millions of riders and drivers who rely on it every day.

###

¹Consumer Reports only shared partial and redacted data that they collected. The data included anonymized rider identifiers, city, original price, discounted price, and information about the discount. The data was grouped by origin and destination. Critically, the data did not allow us to observe trip request timestamps or actual location, or to match these records with Uber’s databases. Consumer Reports did not share any data they collected on driver fares used to calculate the take rate.