Sharing the Road — Travel Efficiency

Jonathan Wang
Uber Under the Hood
8 min readSep 26, 2019


This is one part in a series of sustainability announcements, an overview of which can be found here. For our proposed methodology on measuring Carbon Intensity, see this blog post.

Successful cities are crowded places. Keeping them successful means making better use of their limited physical space by managing travel and minimizing its adverse effects.

It is not a secret that U.S. cities have become overly accommodative of low occupancy personal vehicles. For example, in the United States (US) ~89% of all ground travel vehicle mileage occurs in personal vehicles. Single occupancy car travel represents ~60% of that. Not surprisingly, government agencies like the Department of Transportation (USDOT) and California’s Air Resources Board have worked for decades to reduce vehicle miles traveled (VMT). Despite their efforts trends continue to move in the opposite direction. Given the present situation, how can cities respond to an expected doubling of mobility demand in a world where personal vehicle ownership and use remains the overwhelming preference?

Public roadways are a shared resource. Whether by bus, private automobile, or ridesharing platform, passengers traveling from A to B use some of this resource while moving. More efficient use of public roadways can mean more people getting where they need to go in the time they have. Lower efficiency can lead to painful consequences like unwanted congestion, lost time in traffic, more tailpipe emissions, and the pressure to expand roadways to make more space for cars at the expense of space for people and active modes. Those familiar with the concept of induced demand will surely nod their head here; the effect of this cycle can be seen throughout the U.S.

The second of ten Shared Mobility Principles calls for a “focus on moving people, not cars.’’ At Uber we hope to build new technologies that enable more livable cities by taking approaches that align with the Principles. So, we have a challenge.

To offer on-demand mobility platform competitive with the convenience and affordability of personal car ownership, Uber’s mobility platform requires slack in the form of available drivers awaiting passengers; however, too many empty cars and not enough passenger movement can lead to undesirable conditions for all road users.

Moreover, the broader consequences of inefficient road use directly affect our business. For instance, congestion leads to delays for riders and poor experiences for drivers. Since you can’t manage what you don’t measure, our first question is around measurement. Given the need for on-demand shared mobility, the scarcity of roadway space in our most successful and congested cities and the varied trip purposes and preferences of those who travel, how can we measure the benefits and costs of that travel?

Measuring Travel Efficiency

To answer that question, we introduce the Travel Efficiency metric. In collaboration with transportation consultants at Fehr & Peers, we developed this metric to better understand the productivity of movement on Uber’s platform and enable comparisons across all travel modes. Together with the work we have done in the past on curbside productivity, this extends our ability to learn how to more efficiently utilize public road space. We developed Travel Efficiency with existing measures of vehicle occupancy in mind (e.g.. Average Vehicle Occupancy used by the US Department of Energy and others) and expanded the scope to apply to the shared platform context in which Uber operates.

Travel Efficiency measures how effectively a mode of travel optimizes the movement of people over vehicles. Our proposed method improves existing measures by accounting for deadheading (an informal American trucking industry term, here applied to any situation where a driver is moving a vehicle without passengers) and vehicle relocation in shared-use transportation systems.

Travel Efficiency is defined as a ratio of two measures:

Person miles traveled (PMT): the total distance traveled by each passenger inside a vehicle. In privately owned vehicle travel this figure usually includes the driver as a passenger. For cases where the operator of the vehicle is not traveling to the destination (i.e. general ridesharing, public transit), the operator should be excluded as a passenger. For cases where the operator of the vehicle is traveling to the destination (i.e. carpooling), the operator can be included as a passenger.

Vehicle miles traveled (VMT): the total distance traveled by the vehicle. This includes deadhead miles, the distance traveled where the vehicle does not have any passengers.

Combining these two as a ratio of PMT to VMT provides a simple measure for how efficient a transportation system is at moving people versus vehicles. For example, if a bus followed this sequence of events:

Source: Fehr & Peers, Example for Illustrative Purposes

  1. Left bus yard with driver only and travelled to service area for 2 miles.
  2. Picked up 30 passengers (excluding the bus driver) and drove for 5 miles.
  3. Stopped and dropped off 15 passengers.
  4. The remaining 15 passengers (excluding the bus driver) are driven for another 10 miles and dropped off.
  5. Returned without passengers to bus yard for 3 miles.

Then its Travel Efficiency for this sequence could be computed as:

On-demand, door-to-door ridesharing adds an additional element. Drivers will often have periods where they do not have a passenger in the vehicle but may accrue some deadhead mileage as they travel to new areas where ride demand may increase. The framework for travel efficiency categorizes rideshare drivers’ vehicle mileage into three periods.

  1. Online: Driver has an empty vehicle and is actively awaiting a ride request from their next passenger(s).
  2. En-Route: Driver has accepted their next ride request and drives an empty vehicle to pick up their next passenger(s). This, combined with the Online period, comprise deadhead mileage.
  3. On-Trip: Driver has picked up passengers and is en route to the passenger’s destination.

Source: Fehr & Peers, Example for Illustrative Purposes

Using the above diagram as a reference, we can follow a similar exercise as described above, but assuming the driver picks up two passengers.

  1. Vehicle drives for 3 miles with an empty vehicle to area where the driver expects ride requests.
  2. Driver receives a ride request from passengers 2 miles away.
  3. Driver picks up 2 passengers and drives 5 miles to their destination and drops them off.

The Travel Efficiency for this sequence would then be:

This ridesharing example is fairly straightforward, but there are some additional unsolved complications for ridesharing platforms in particular which may bias estimating Travel Efficiency.

  1. Double counting vehicle distances traveled across platforms. In most markets, there are multiple platforms for which drivers can be simultaneously operating on to provide passenger mobility or food and parcel delivery services. It is possible for one platform to record mileage from a driver as “on-trip” while another captures the same data as “open.” Since independent platform operators do not have visibility into competitors’ systems, these miles will be double-counted and Travel Efficiency biased downwards.
  2. Difficulty in estimating driver off-trip mileage purpose. Drivers operating on ridesharing platforms will sometimes turn off the app and drive to areas that they believe to have higher demand and thus result in a higher likelihood of getting a ride request. Drivers can also leave the app on while off-trip and be driving to a destination for their own purposes. Shared mobility companies are generally unable to observe the purpose for which off-trip mileage is being generated. This could both bias calculating Travel Efficiency upwards or downwards.
  3. Adjusting for “wobble” in shared trip modes. Pooled trips can require detours in order to pickup and dropoff additional passengers. The PMT generated should be adjusted downward to the theoretic most efficient way to complete each leg of the trip while using the actual generated VMT.

With these caveats for ridesharing in mind, utilizing the Travel Efficiency metric can be an intuitive method of assessing how effective a mode of travel is at moving people. Now that we have defined a metric, we address how the metric should be used to compare across modes.

Moving Forward

The Travel Efficiency metric provides a baseline framework for assessing the ability for a mode of travel to move people with fewer vehicles. As we continue to better understand the role that ridesharing plays in the overall transportation ecosystem, there are natural extensions and follow-ups to this methodology, including:

  • Measure how Travel Efficiency varies across modes and other factors. Travel Efficiency can be measured for other modes of transportation as well as how Travel Efficiency can vary across different trip purposes, time of day, and other factors.
  • Measuring per passenger mile carbon emissions. Integrating vehicle mileage data with Travel Efficiency enables calculating average per passenger mile carbon emissions of a mode of transportation.
  • Measuring micro-mobility platforms. While micro-mobility options such as scooters and e-bikes don’t generate vehicle miles in the same way a car does, they can also be integrated by applying a Passenger Car Equivalent adjustment factor. Additionally, the mileage generated by service vehicles used to reposition, maintain and recharge these devices can also be measured.
  • Apply Travel Efficiency to goods movement. This methodology can be extended to evaluate the ability for a transportation mode to move prepared food, groceries, parcels, freight and other goods.
  • Assessing Travel Efficiency for Transportation Ecosystems. No mode of transportation exists in a vacuum. The Travel Efficiency metric can be applied more holistically in evaluating the efficiency of travel across all available options to accomplish a set of trips. This would enable direct comparisons between single mode, multimodal, and shared modes of completing trips. An example of this is comparing the Travel Efficiency for a commute between driving a personal vehicle for the first mile to a transit station versus taking a pooled rideshare.

The goal of this work is to provide an updated set of tools for evaluating the efficiency of different transportation modes that are inclusive of new mobility options. We also want to highlight the importance of accounting for context when evaluating and comparing between modes.

In addition to working on these extensions, we are developing solutions to improve the accuracy of measuring these figures, and product innovations that can improve overall efficiency performance through better integration with the transportation ecosystem in cities. We are excited to continue to study the role that ridesharing can play to strengthen mobility in cities.

We welcome your thoughts and feedback on the above proposed approach. Please contact us at to provide input.



Jonathan Wang
Uber Under the Hood

Policy Research Scientist @Uber, Executive Director @Deltanalytics