Transportation Terms: Dwell Time
My first installment for a new series I’m calling transportation terms, is ‘dwell time’ [1]. Dwell time is the amount of time that a vehicle spends at stops or stations, as opposed to en route between them. It is important for two reasons. First, for an urban bus or train with frequent stops— and these are the two transportation modes to which this metric is most frequently applied — the time spent stationary, picking passengers up and dropping them off, can be a significant percentage of total travel time, particularly at stations with high passenger demand, on systems that are running at high utilization. Research from McGill University [PDF] found it to be approximately 10–25% of a typical bus journey [2]. The primary reason is that most public buses, unlike subways and light-rail, only allow boarding from the front of the bus, and require the driver to verify each rider’s fare as well as to sell tickets to any rider who wants to buy one on board. Since drivers can’t get back on the road until all passengers have boarded and had their fares checked (or sold), it’s not hard to see how a stop with ten or fifteen people waiting to board could really slow things down. The most common solutions for speeding up bus dwell times are those that everyone who has ever ridden the train knows: off-board fare purchase and all-door boarding. Indeed, I first came across the term “dwell time” in a recent Streetsblog USA article, All-Door Boarding Can Save Time for Bus Riders — If Transit Agencies Embrace It.
The second reason dwell time is so important is that people hate it. People want to feel in control , and dwell time is like the transit equivalent of traffic. Surveys have shown that the average rider who has to wait for people to push their way on and off of packed subway cars; for the bus driver to take tickets; etc.; or in fact, for anything beyond a quick stop for people to get on and off, will blame the transit agency. If you can reduce dwell time, it follows, you can get a disproportionate increase in satisfaction. In the McGill study referenced above, researchers found that 67% of riders who used the middle or rear doors perceived an average reduction in travel time. Particularly notable is that while the move to all-door boarding did produce an observable reduction in travel time, the perceived reduction in travel time was 25% greater than that which had actually occurred. In other words, every dollar that was invested in this project got free leverage of 25% in terms of customer perception.
The tech angle:
Cashless payments are one of the more subtly brilliant features Travis Kalanick incorporated into Uber’s product from the very beginning. In the original vision of Uber as “everyone’s private driver,” cashless payments seems to have been designed as part of the “luxury” experience; does anyone with a private driver pull out their wallet at the end of a ride and fumble for change? But it was more than simply feeling like a baller that made cashless payments so great. Consciously or not, Travis had landed on an elegant solution to one of the primary causes of that hated problem, dwell time.
Back when I would take NYC yellow cabs [3], the experience of taking one in peak hours often went something like this:
- Run to a cab pulling over in your general vicinity (15 seconds);
- Realize that there are passengers getting out; wait for them to pay and exit. Keep scanning for other taxis that may be more readily available (60 seconds, but could be much worse if passengers are tourists. I’ve definitely given up on more than a few car fulls of shoppers trying to figure out who is paying and how they are going to get their shopping bags out);
- Get into car ( 5 seconds);
- Tell the driver where you are going as he waits in order to figure out which lane and direction to go (15 seconds);
- Drive to your destination (15 minutes);
- Fumble for your wallet and pay (45 seconds);
- Get out of car (10 seconds).
If you do the math, you’ll see that dwell time (steps 2, 3, 4, 6 and 7) add up to 2.25 minutes. That’s almost 15% of the 17.25 minutes (2.25 + 15) that your total journey takes. Worse, dwell time for a taxi is fixed, regardless of the length of the trip since there are no intermediate stops, so the shorter the trip, the higher the % of it that is dwell time. If the ride was only 10 minutes, dwell time would be nearly 20% of the journey. With cashless payments, in one stroke Uber has eliminated the payment step and reduced the total activity that needs to happen at alighting to the few seconds it takes to get out of the car. Let’s say that it takes 5 seconds to get in a car and 10 seconds to get out, conservatively. That reduces dwell time above to 45 seconds, or 5% and 7% of the journeys above, respectively; and that is without any of Uber’s other features.
In reality Uber does a few things that reduce dwell time even further. First, online passenger/driver matching ensures that there never is a previous passenger for whom you have to wait to alight before you can get in. That saves the 15 seconds that was left of waiting for them, even under a cashless payments system. Now we are down to 0.5 minutes, or 3% and 5% of journey time for a 15 and 10 minute drive, respectively. Uber added the ability — and later made it a requirement — to input your destination before you e-hail the car, eliminating the 15 seconds where you tell the driver where to go and he enters it into his GPS, figures out which direction to start in, etc. This takes dwell time down to 0.25 minutes, or 1.5% and 2.5% of journey time, respectively [4].
The features that Uber introduced to the for-hire vehicle world — cashless payments, mobile e-hails, dynamic passenger/driver matching, pre-boarding destination entry, and dynamic routing — have become standard among a new generation of app-enabled transportation startups; they are now merely the cost of admission for playing in this space. With the rise of multi-passenger “transit-like” transportation services such as Bridj, Via, Lyft Line, and Uber’s own UberPOOL, and these companies’ growth projections — not to mention the expected explosion in service when autonomous vehicle technology is finally mature enough to allow them to procure fully driverless for-hire vehicles — it will be interesting to see what happens to dwell time in this context. These services have dwell time in the traditional sense of the term: time spent at a stop where one passenger waits for another to board or alight. Many have taken the additional step of putting some of the responsibility for minimizing dwell time on their riders, charging if you are not on board within a certain amount of time after the vehicle arrives at the pickup location, and/or simply leaving the pickup location, and you, when this time has passed [5]. This is unprecedented, but it is also a feature that was previously impossible and only makes sense in the current paradigm: fixed-route public transit modes could never do this, since they do not provide door-to-door service, nor do they wait for specific riders; taxis could never do this, because for the most part they do not have the rider’s payment info in advance, and by nature of the highly fragmented taxi industry, anyone who keeps their payment info on file is probably a big repeat customer, not worth antagonizing. My own opinion is that it’s a great idea, as it gives riders a sense of agency over their ride, something that encourages better behavior as well as retention. If one service thinks that the dwell time standards required of riders by other services are too draconian, they can always relax theirs as a competitive advantage (though they then risk an adverse selection problem where they become the preferred service of chronically late riders and end up chronically late themselves).
I suspect that many of the innovations in transportation technology do or will address dwell time in one manner or another.
We might see another application of dwell time to TNCs and other for-hire vehicles in the not too distant future: as a metric used for calculating curb throughput/utilization and its impact on traffic. As the need for parking transitions to a need for passenger pick-up and drop-off zones, gradually at first as more and more trips mode shift from personal vehicles to e-ride-hailing, and then suddenly with the diffusion of truly driverless cars, cities will need to ensure that these zones — referred to as SUMZ, or SUM Zones, by Greg Rogers and Patrick Smith, short for Shared-Use Mobility Zones — are used in an efficient, orderly, and equitable manner. Otherwise any traffic reduction from people no longer needing to circle for parking will only be replaced by the traffic created by double-parked vehicles waiting for passengers to board and alight (this type of traffic already exists in the denser parts of New York City, or anywhere that a taxi or Uber stops to pick up a passenger on a narrow street).
Cities will need enough SUM zones, each with sufficient throughput, respectively, so that shared-use vehicles do not form lines and stick out into the street’s thru-traffic lanes (including bike lanes). A high throughput will require mobility providers to keep their dwell time low; dwell time and frequency will be important metrics to predict whether this is in danger of transpiring.
If queues do start to form, one of the city’s first steps will be to do a side-by-side comparison of dwell time by service provider and identify the companies whose vehicles are lounging in the SUM Zones the longest. The city will push any laggards to lower their numbers; if the providers do not comply, the city can simple deny that provider access to the SUM zone(s) or levy fines for each use.
[1] I’ve been thinking about adding new “features” to my Medium blog for some time, as a way to post more frequently as well as not have to come up with smart things to say every time I want to post (“Features” sounds overly pompous to me. But I can’t think of a better word. Suggestions?). As my first such feature “Transportation Terms” serves the dual purpose of helping me remember the new vocabulary I learn as I get deeper in the transportation world, and of challenging me to explain this vocabulary in a clear manner to others, guided by the philosophy of, “if you really want to learn something, teach it.” I’ll be introducing new terms only as frequently — or infrequently — as I learn them myself, and will share the context in which I do so. Because I bring a technology background with me to my discussion of cities and transportation, I will also try to tie each term to its application to the transportation technology world.
[2]Subways and light-rail with high utilization and busy platforms suffer from the same issue. Imagine a subway system with stops spaced 5–10 minutes of travel time apart. Now imagine a subway car on that system with passengers pressed together like sardines, standing room only, like those in DC or NY at rush hour, such that people getting on and off the train have to push their way through the crowds, holding the train up while they pass. If the average travel time between stations is five minutes, the average dwell time only has to be 60 seconds—ie the train be in the station for an average of 60 seconds to let passengers on and off —for dwell time to be 17% of your trip! This is why the MTA is constantly making announcements in the subway cars to “Please step aside so that passengers may enter and exit the train [tk]” and “Please use all of the car” (most people’s natural inclination upon boarding, if seats are no available, is to stand in the area near the door. Unfortunately when this area fills up it becomes harder — ie takes longer — for additional passengers to board. When passengers move to the middle of the car, they free up space by the door, allowing boarding to happen more quickly).
[3] It feels weird to say it, but for an Uber user taxis are the new land lines. My hired car usage has shifted so completely from taxis to Uber/Lyft that the only way I can accurately describe riding in them is in the past tense.
[4] These calculations all treat the trip as starting when the passenger makes contact with the vehicle. If you were to include the time that the passenger spent hailing the vehicle (in the case of a taxi) or the time between the hail and the vehicle’s arrival (in the case of a TNC vehicle like Uber) as part of dwell time, then it would be a much higher percentage of total journey time.
[5] Uber even does this for single-passenger trips, charging passengers who make drivers wait more than two minutes or take more than two minutes to cancel. This makes sense for them, as their drivers are not employees they can order around but in a sense are customers of theirs almost as much as the riders.
Thanks to The Overhead Wire for reviewing earlier versions of this post, and in particular for the idea of applying dwell time to curb throughput and utilization.
