Dynamic Pricing by Neighborhood

A Closer Look at San Francisco

Previously, we examined general ride-sharing prices across entire cities. Next we turn our attention to how pricing varies geographically across individual neighborhoods in San Francisco. You've probably observed that Surge and Prime Time vary based on where you're getting picked up; during the World Series you will likely encounter high Surge pricing in SOMA while your friend in the Mission is being quoted a regular fare.

We created an interactive map showing how frequently Surge and Prime Time appear in each neighborhood. Darker tones indicate a higher likelihood of encountering each. We’ve also listed the average pricing multiplier within neighborhoods to represent the severity of Surge and Prime Time. You can explore the data across several time periods: Morning Commute, Evening Commute, Weekday Evenings, Weekend Daytime, and Weekend Nightlife.

Explore the full-page map for the best experience, but you can see an embedded version below.

Dynamic Pricing Frequency In SF Neighborhoods

We can observe that, as expected, morning and evening commutes have opposite dynamic pricing patterns. In the morning, people from outer neighborhoods need to go downtown, so the demand in those areas is higher than normal. Most rides then terminate downtown, so it’s likely that driver supply ends up higher there, resulting in less frequent and severe dynamic pricing in that area. In the evening we see the opposite effect, with higher Surge and Prime Time frequency in downtown and less in residential neighborhoods.

Less intuitive is that on weekday nights after 7:00 pm, the frequency of dynamic pricing is far higher than we would have expected. Over 50% of requests in the central part of the city trigger dynamic pricing, often at high multipliers. We experienced this steep pricing personally when leaving work at 8:30pm on quiet weeknights, and this was one of the motivations for building What’s the Fare in the first place. A possible explanation is that while these are off-peak hours for demand, they are perhaps even more off-peak for drivers who perceive these hours as slow or less desirable. In the future, we believe that improved pricing transparency will allow drivers to discover regions and time periods that are unrealized opportunities and even out bumpy pricing patterns.

Results shown here are from requests to whatsthefare.com from September 8, 2014 to October 16, 2014. Note that we group results by neighborhood because that's the way that most people intuitively understand the city, but Uber and Lyft do not necessarily calculate dynamic pricing across the same geography and it's unclear how they set boundaries for Surge and Prime Time regions. Neighborhood boundaries used here are taken from DataSF. Note that data is sparse for some neighborhoods for the time periods we examined, and these are colored in light gray.

It Pays to Compare

Ride-sharing by the Numbers

- See how different ride-sharing services measure up against each other and against taxis.

A few weeks ago we built What's The Fare as a personal side project because it was frustrating to switch between multiple apps to check current pricing and to estimate fares. Several weeks after launch and hundreds of thousands of estimates later, it's clear that many other users feel the same way.

While our goal remains to simply display accurate fare estimates across services to users looking for rides, we also believe that ride-sharing customers can benefit from understanding the general pricing trends of these services.

We found that Uber, Lyft, and Sidecar provide a cheaper alternative than taxis in the vast majority of cases, but there are predictable times where dynamic pricing (i.e. Surge and Prime Time) is not the exception but the rule. In these cases, the decision to take a cab or a ride-share isn't always clear. Further, price-sensitive riders can achieve significant savings by checking the best-priced service for each ride rather than blindly sticking to any one service.

We're in a fortunate position to study this because we gather real-time pricing information with each user request on What's The Fare, allowing us to compare estimated prices across multiple services for the exact same route. While it's important to note that the resulting prices are just estimates and not actual fares paid, the estimates do include the services' real-time availability of drivers and dynamic pricing multipliers if applicable. For additional detail on methodology and caveats see below.

Today, we're sharing a few of the most interesting findings from the highest-traffic cities of San Francisco, New York City, and Los Angeles based on several weeks of data from mid-September through the first week October. Let's get started with the basics.

Are Ride-Sharing Services Actually Cheaper Than Taxis?

There's been some debate about whether ride-sharing services are more affordable than taxis after factoring in dynamic pricing (Surge and Prime Time) and availability. Looking at fare estimates from UberX, Lyft, and Sidecar, we wanted to see how often one or more of these services was cheaper than a taxi.

Ride-sharing Delivers

The results are quite clear. In the overwhelming majority of cases, even with dynamic pricing at least one ride-sharing service offers a lower price than a taxi almost all the time in San Francisco and Los Angeles, and 85% of the time in New York.

The estimates in this chart don't take into account how long you would have to wait to get a ride. We discuss ETAs in more depth below.

Percent of Rides With Lowest Fare
San Francisco
New York City
Los Angeles
How Individual Services Compare

Things get more interesting when you compare individual services to taxis, however. While collectively ride-sharing services are almost always cheaper than taxis, individually there's more variability. The graphs below show the percentage of time that each service is cheaper than a local taxi. In Los Angeles any of the ride-sharing services are cheaper than taxis over 95% of the time. But in San Francisco and New York, UberX only winds up cheaper than a taxi three out of four rides, and Lyft is cheaper 72% and 64% of the time, respectively.

Note that Sidecar is currently not available in New York City.

Percent of Rides Cheaper Than a Taxi
San Francisco
New York City
Los Angeles
What's the Deal with Dynamic Pricing?

Regardless of your position on the controversial topic of Surge and Prime Time pricing, it's hard to say it's not an interesting subject. Proponents argue that dynamic pricing gets more cars on the road and increases reliability, while critics claim it creates an inconsistent experience and gouges consumers. While we don't have evidence to conclusively argue for or against either side, we can present data on the frequency and magnitude of dynamic pricing.

The Rush Hour Premium

It's not at all surprising that there's greater demand for rides during rush hour, but just how often are Uber and Lyft using Surge and Prime Time pricing during these periods? Below we've graphed the percentage of estimates per hour that were subject to dynamic pricing during the work week.

It's not looking great for commuters in San Francisco, which sees the most extreme price fluctuations during commute hours. At 8am and 6pm, you are in fact more likely than not to be hit with dynamic pricing. New York and Los Angeles both show commute-time peaks, but at significantly lower magnitudes. However, for rides beginning and ending in Manhattan specifically, we do see a significant 8am commute spike, and also the only significant example of Surge for Uber Black that we've discovered thus far.

Percent of Rides With Dynamic Pricing - Weekdays
San Francisco
New York City
Los Angeles
Not All Surges Are Created Equal

Other than frequency, the other important number to look at is the actual dynamic pricing multiplier. After all, a 25% Prime Time tip is quite a bit more affordable than 4.0x Surge. The following graphs show Surge and Prime Time values as a percentage of all fare estimates across the different days of the week. Put simply, darker colors mean progressively higher fare multipliers and thicker bands mean more occurrences.

While dynamic pricing is often portrayed as an infrequent occurrence during busy periods, we see from these charts that it's actually in effect very often. For example, in San Francisco on weekends you're more likely than not to see dynamic pricing. Lyft's Prime Time seems to be in effect more often overall than UberX's Surge. Surge and Prime Time aim to exchange reliable pricing for reliable availability, but it's worth noting that dynamic pricing is used far more often than most would expect.

It's worth pointing out that Uber and Lyft use different scales to describe their dynamic pricing. 1.5x Surge is equal to 50% Prime Time, and a 3.0x Surge is equal to 200% Prime Time.

Percent of Rides at Each Dynamic Pricing Level
UberX San Francisco
Lyft San Francisco
UberX New York City
Lyft New York City
UberX Los Angeles
Lyft Los Angeles

From the above charts, it appears that UberX and Lyft are roughly following the same trends in terms of aggregate frequency of dynamic pricing. Now let's look deeper at how closely they track each other for individual fare requests.

Are Surge and Prime Time Highly Correlated?

Anecdotally, many users who encounter dynamic pricing in one ride-sharing app will check a second or third app and often see a different multiplier. We wanted to investigate whether UberX and Lyft are likely to apply similar dynamic pricing for individual rides. For the same route at the same time, how likely will it be that their multipliers are the same, or at least similar?

To find out, we plotted the Surge and Prime Time multipliers in effect for each request. In the charts below, the horizontal axis is the UberX Surge multiplier, while the vertical axis is the Lyft Prime Time multiplier. For example, the bubble at 1.5x horizontally and +25% vertically shows the number of times that UberX reports a 1.5x Surge multiplier at the same time that Lyft reports a +25% Prime Time increase. The size of each bubble represents the relative number of occurrences.

Correlation Between UberX and Lyft Dynamic Pricing
San Francisco
New York City
Los Angeles

In a world where dynamic pricing reflects overall passenger demand for transportation, we'd expect dynamic pricing from UberX and Lyft to be reasonably correlated. The circles along the diagonal from lower left to upper right would be larger than those along the left and bottom edges.

In practice, we see the left and bottom edges account for more occurrences than the diagonal. This shows that Uber's Surge and Lyft's Prime Time are not as correlated as one would expect. Indeed the correlation coefficient is only 0.44 for San Francisco and even lower for New York and Los Angeles at 0.31 and 0.38. This may suggest that dynamic pricing is currently more a function of each service's individual driver supply than of overall passenger demand.

Since dynamic pricing isn't strongly correlated, passengers will save if they check multiple apps in attempt to avoid dynamic pricing. We discuss the hypothetical savings of comparing across services versus being loyal to a single service below.

Which Service Should I Use?

We've learned that in general ride-sharing services are cheaper than taxis, and that UberX and Lyft are using dynamic pricing regularly. You might be thinking "just tell me which service to use already". Unfortunately, even for the reader who only cares about price, the response is "it depends". There isn't a hands down winner in any market, and the answer changes frequently. For users who care about getting a ride quickly, the answer gets even more complicated.

Cheapest Service By City

We've graphed what percentage of the time each service is the cheapest. In San Francisco, no service is cheaper a majority of the time. Sidecar and UberX are cheapest most often, about a third of the time each, followed by Lyft at about 19%, and the rest of the time either multiple services offer the same fare or taxis win out.

In New York and Los Angeles, Lyft either has the lowest fare outright or is tied for the lowest fare about half the time. But it's important to note that Lyft and UberX have set their base, per-minute, and per-mile prices identically in these two cities, so Lyft wins most of these fares only by a slight margin, since they round down to the dollar. Otherwise, Lyft and UberX only differ when dynamic pricing is applied at different multipliers. What's clear is that there's a lot of variability across all markets as to who is cheapest and how often.

To be more realistic we only considered fares from ride-sharing services when ETAs were 7 minutes or lower. While arbitrary, we believe this is a reasonable approximation for how long many users are willing to wait in metropolitan areas, though we recognize that for sprawling areas like LA the acceptable wait times may be longer. As we don't currently have ETA data for taxis, for the purposes of this chart we simply assumed that a taxi would always be available.

Percent of Rides With Cheapest Fare
San Francisco
New York City
Los Angeles

In addition, these graphs don't reflect what happens as the different services make price cuts. For example, both UberX and Lyft had generous summer discounts in San Francisco. The analysis above covers dates that coincided with UberX's latest price cut on September 15 through the present. When we looked at SF estimates from the week prior, Lyft was cheaper than UberX the majority of the time. The most affordable option can change week to week with little notice. Savvy consumers will make sure to note price cuts and choose services accordingly.

SF Before UberX Price Cut
SF After UberX Price Cut

Sidecar was omitted from these charts to highlight the effect of the Uber and Lyft price promotions.

Cost of Loyalty

Given that there isn't a clear winner in any market, we decided to calculate "the cost of loyalty". In other words, what is the additional cost of always choosing to take rides from just a single provider rather than shopping around for the best price before every ride? Naturally, the results vary quite a bit by city, but what is consistent is that the savings are significant if you compare prices instead of sticking to a single service.

In San Francisco and Los Angeles, if you are loyal to just a single service, Sidecar is your best bet. It should be noted though that Sidecar's driver ETAs are the longest of the ride-sharing services we looked at (more detail below), so you will likely have to wait for rides longer in order to realize these savings. If you are loyal to UberX or Lyft in San Francisco or Los Angeles, expect to pay 14-43% more over the long term than if you compare before you ride. In both these cities, even with frequent dynamic pricing, it is still far cheaper to stick with one of these services than always calling for a taxi.

New York, and Manhattan especially, buck the trend. This is likely because UberX and Lyft base prices are set fairly close to taxi prices, which are already lower than taxi prices in San Francisco and Los Angeles. As a result, just a small amount of dynamic pricing can make UberX and Lyft more expensive than taxis. We see the fascinating result that if you're going to be loyal to any one service in New York, taking a taxi every time will save you more than taking UberX or Lyft every time and putting up with dynamic pricing. In Manhattan, always taking a taxi would seem to cost you only about 6% more than comparison shopping, assuming you take rides at a variety of days and times when ride-share services are subject to dynamic pricing.

This metric is difficult to estimate. Our methodology was to compare the cumulative costs for each service across thousands of user requests. We only considered requests that had estimates for all services, in order to avoid inadvertently biasing towards services that that may have been lacking an estimate. You can think of it as, if I were to take 1,000 rides over my lifetime with one individual service, how much more would I pay than if I compared prices and always picked the cheapest option.

Cost of Loyalty Over Always Choosing Cheapest Fare
San Francisco
New York
Los Angeles
Don't Want to Wait?

We don't think price is the only important factor to consider when choosing which ride-sharing service to use. In the real world, wait time is often equally or more important than price.

In our final analysis, we looked at the ETAs of the nearest available drivers by graphing the cumulative distribution of ETAs by service. The horizontal axis is the ETA in minutes and the vertical axis represents what percentage of rides are available within that ETA.

ETAs in this analysis are ETAs reported by the ride-sharing services. By definition, ETAs are only estimates and without booking rides there's no way to verify that actual arrival times are the same as the estimates.

Overall, San Francisco is the city where users can expect to wait the least. There, UberX and Lyft are virtually indistiguishable, with both services showing ETAs of 5 minutes or less for more than 85% of rides. Sidecar lags a few percentage points behind but is still competitive. It should be noted though that for Sidecar this shows the ETA of the nearest available driver and not necessarily the driver offering the lowest fare; the ETAs of those drivers on average are noticeably longer. Similarly, there is no guarantee that Uber or Lyft will assign you to the nearest driver, so actual wait times on average may indeed be higher in practice.

In Los Angeles, a much larger city geographically, ETAs are unsurprisingly longer. Lyft has the best ETAs, followed by UberX, and both have a driver within 7 minutes more than 75% of the time. Sidecar only has a driver within 7 minutes 40% of the time.

It's interesting to note that Sidecar is the service with both the best cost of loyalty but the longest ETAs in the cities in which it's offered. Essentially, Sidecar allows you pay lower prices in exchange for longer wait times.

Percent of Rides Available Within an ETA
San Francisco
New York
Los Angeles

As we've seen, the question of which service to use is complicated to answer. It depends on your city as well as the time of day, the day of the week, and how long you're willing to wait. Even if there was a more definitive answer, it likely wouldn't be valid for very long since pricing changes often, so we encourage you to closely watch for price cuts and avoid dynamic pricing to get the best fares. In fact, it was the uncertainty around trying to answer this question before each ride that motivated us to build What's the Fare in the first place.

Until Next Time

We hope this adds valuable data to the discussion around ride-sharing, and we welcome questions, clarifications, and analysis requests. In the coming weeks, we’re excited about collecting more real-time fare estimates and sharing additional findings, including an exploration of UberPool, Lyft Line, and Sidecar Shared Rides.

If you're curious to see data for your city, please help spread the word about What's The Fare. The more users we have in a city the more data we'll be able to use in our analyses.

- Jonathan and Matt

About Us

Jonathan and Matt have worked together for years building products, and combined have led engineering and product teams at YouTube, Google, Qwiki, and Bonobos. What's The Fare is a project built from their shared interest in ride-sharing and transportation.

Follow us at @whatsthefare, @jon_goldman, and @matthewliu. Or, email us at info@whatsthefare.com.

Further Research

For data geeks who are interested in a bunch of graphs we didn't have room for here, as well date ranges and sample sizes, we've compiled a full post for each city: San Francisco, New York, Manhattan, and Los Angeles.

Methodology and Caveats
  • We are not in any way affiliated with Uber, Lyft, Sidecar, or any taxi providers.
  • All findings are based on user-initiated fare requests from whatsthefare.com, and not from any automation or scraping.
  • Uber, Lyft, and taxi fares are computed using each service's publicly posted rates, applied to a formula using the distance and duration of the most likely driving route according to Google Maps. Lyft rounds fares down to the whole dollar, while Uber does not. Realtime dynamic pricing based on the pickup location is applied as appropriate as indicated by the services themselves. Traffic conditions are not considered.
  • Estimates do include the "safe rides fee" where applicable, and New York taxi estimates include time-of-day and state tax surcharges. Estimates do not include tolls or any other potential surcharges. However, in most cases, tolls do not even apply, for example, for all San Francisco data here, rides originate in the city of San Francisco which means any bridge crossings in that direction would not be subject to tolls. Since all estimates for Uber, Lyft, and taxis are computed without tolls, the comparisons are apples-to-apples in this respect, so while including tolls could affect the percentage difference between fares, it would not change whether one service was more expensive than another.
  • Taxis use a different pricing scheme than Uber and Lyft. Uber and Lyft charge a base fare, plus a distance component based on the full distance traveled, plus a time component based on the full duration of the ride. Taxis charge a base fare and then switch back and forth between charging based on distance, when the vehicle is moving faster than a certain speed, and based on time, when the vehicle is moving slower than a certain speed. This means that taxi fares are more difficult to estimate and more likely to vary based on conditions.
  • All prices for taxis include a 15% tip in order to simulate the out-of-pocket price you're likely to actually pay.
  • Sidecar is priced differently than the other services, and provides a choice of guaranteed fares for each ride. Fares shown for Sidecar are therefore the fixed fares as obtained at the time of the request, and not a time-and-distance estimate as with the other services. Unless otherwise noted, the Sidecar fare used in these analyses is the best fare where the driver ETA was at most 7 minutes, if any, and otherwise the best remaining fare.
  • Requests selected for each region's analysis were those where the pickup point originated in San Francisco County, the 5 boroughs of New York as indicated by each borough's county boundaries, and Los Angeles County, with boundaries coming from US Census TIGER data. Only rides of less than 100km driving distance were considered. The sample size and date range for each graph can be seen on the city-specific graph summary pages: San Francisco, New York, Manhattan, Los Angeles.