Precarious by Design

United Kingdom

9 October 2018

A notification: “Meal, 2 min away.” My phone is beeping and flashing, urging me to accept the opportunity with the help of a 15-second countdown bar. “Hurry up!” the indicator seems to say. “There are plenty of others poised to take on this gig.”

It’s my second weekend of cycling as a delivery partner with UberEATS in London and I’m beginning to get the hang of the Uber Driver app, a piece of interface design that mediates every dimension of the job. That job, I should declare, is not what I do for a living. I am on a full-time contract with the publication you’re holding and a beneficiary of the existential securities that come with employee status in the United Kingdom. The reason I’ve signed up with UberEATS, which offers very few protections, is to understand the role played by interface design in the ballooning on-demand economy. And while my day job may appear secure, I am acutely aware of the increasing gigification of my line of work, with setups such as Fiverr and TaskRabbit now offering copywriting, editing and translation services for end-user rates as low as $5 per job. This new economic model affects me and it will almost certainly affect you. My most immediate concern right now, however, is my flashing phone: the countdown bar is ticking; above it a map indicates a two-minute route I can take to the restaurant. It’s just around the corner, so I tap the screen to accept the job and set off.

UberEATS is a branch of Uber, the global enterprise that launched as UberCab in San Francisco in 2009. In 2011, the company dropped its “Cab” suffix after complaints from San Francisco taxi operators. “Cab” implied that Uber was a cab service, the authorities reasonably argued, and this meant it should comply with local taxi regulations. Uber didn’t and still doesn’t. Since dropping the suffix, Uber has described itself as a tech company rather than a transportation or delivery service: it simply provides a platform, it claims, on which independent contractors can solicit jobs. For use of this platform, the company extracts what it argues is a reasonable cut: between 25 and 35 per cent per job. The platform for both Uber and UberEATS consists of the Driver app (which is what I am using), the end-user apps (which are the interfaces through which you book either a car or some food) and the inscrutable cocktail of algorithms that link the two together. It is because of the company’s role in pioneering this form of business model that Uber has become known as poster child of the gig economy.

On this weekend in May 2018, I am using a Driver app that has been subject to numerous accumulative redesigns since 2010. Only a month ago, however, on 10 April, Uber announced that it had designed a brand-new Driver app entirely from scratch. Most driver and delivery partners haven’t been notified of the design overhaul yet, but it is set to be slowly rolled out across the hundreds of cities that Uber operates in over the coming three or four months, and more than 3 million people will potentially use it. Now seems the right time to assess the affordances of the current Driver app and what the redesign professes to achieve.

My pick-up point is a pizzeria. I secure my bike, make myself known to the staff, and show them the order number on my phone. I’ve been advised in a chirpy instruction video (“Getting started with UberEATS” on Vimeo, my only formal training for the job) to separate hot and cold foods in my box backpack and always to double-check the order, the complete details of which I can view in the app (“Incomplete orders are very frustrating to the eater,” the video’s narrator warns).1 It is only once I’ve verified these and swiped “Start trip”, that my drop-off address is revealed. This bears repeating because it comes as a surprise to many Uber end-users: cyclists and drivers have no idea what their final destination is when they accept a job, or what the fare will be for that matter. This is known as “blind driving” and is integral to Uber’s strategy – it means there’s no destination-based bias on the part of the driver or cyclist, and that passengers and “eaters” can always rely on securing the service with a few taps of their thumb. This structural feature will remain the same in the new Driver app, says Uber spokesperson Michael Amodeo: “We want to make sure that people can get a ride from wherever they are to wherever they want to go. That’s an important principle for us in terms of expanding access.”

Blind driving is good for the end-user, but it can have unwelcome consequences for drivers, as Alex Rosenblat and Luke Stark found in a 2016 study published in the International Journal of Communication. “You’re driving around blind,” says one of Rosenblat and Stark’s interviewees. “When it does ping, you might drive 15 minutes to drive someone half a mile. There’s no money in it at that point, especially in my SUV.” The UberPOOL feature, whereby several passengers going in the same direction are linked on the platform and given a cheaper rate by sharing a trip, is notoriously unprofitable for drivers (“UberPOOR they call it,” says an interviewee in journalist James Bloodworth’s 2018 book Hired: Six Months Undercover in Low-Wage Britain).2 Turning down trips, however, by letting the timer run out or tapping an X labelled “No thanks”, eventually leads to being bumped off the app. If a driver consistently rejects UberPOOL trips, for instance, the app will either log them out for period of time, or, it would appear, implement other punitive tactics: “So it looks like Uber has put me in some sort of time-out, ever since I started turning down every single pool ride for the past month,” writes a Seattle-based driver in the forum “The low quality rides I’m getting from Uber now is noticeable.”

As a driver or cyclist partnering with Uber, you are supposedly self-employed. According to, this means you should be able to “decide what work [you] do and when, where or how to do it”. But blind driving means you’re in no position to see, let alone negotiate, the terms of the jobs you take on. What’s more, you are penalised for acting on the small amount of wiggle room allowed by the choice architecture of the app. “The fact that you have blind routes within an app is an indicating factor that the person [using it] is not running this relationship as a business,” says Emma Wilkinson, an employment law specialist at the charity Citizens Advice. “The problem is, it’s not the be-all-and-end-all. It’s an indicating factor rather than the indicating factor.” Wilkinson says the picture is complicated by a legal precedent which classifies black cab drivers as independent contractors. Under the taxi rank rule, black cab drivers are technically not allowed to turn down a fare, provided it is “reasonable”. “The taxi rank rule,” says Wilkinson, “could equate to this blind driver.” Blind driving seems to sit in somewhat of a legal grey area in terms of employment status. To Ron, one of Rosenblat and Stark’s interviewees, however, it is the main sticking point: “Show the destination before. If we’re independent contractors, we should have the right to refuse.”

My drop-off address is just under two miles away. This seems to be the average distance that the Driver app will send me; eaters can only order from restaurants within a certain radius of their location, which means that orders can be delivered by bike in a reasonable amount of time. It’s a little different for drivers, who have no restrictions as to how far the app can ask them to go. Last year, Uber introduced a “Long trip” notice, allowing drivers to see the distance, but not the direction or destination of a job, before accepting it. I ponder this as I pedal towards the drop-off, remembering an Uber cab ride I had taken from Gatwick airport to my home in London a few weeks earlier. Immediately after accepting the trip, the driver rang me; the interface allows for encrypted in-app calls once a trip is active. “I just wanna check you’re not headed for Brighton,” he explained. Brighton and London are, of course, equidistant from Gatwick but in opposite directions.

The airport driver may have used up his quota of a rationed in-app “Set for destination” feature, which was introduced by Uber in 2015. This allows drivers to set the general direction of travel twice a day (typically for heading out and then home again). Last year, after the departure of Uber’s scandal-swept founder and CEO Travis Kalanick, the company attempted to woo its US and Canadian drivers with a campaign titled “180 Days of Change”. One of its concessions was to allow drivers to use the “Set for destination” button up to six times a day. They loved it, and used the feature so much that it ended up changing the overall market conditions. The company swiftly retracted the move, leaving drivers not a little miffed. As I delve into a residential area in south London, and calculate that I’m about 40 minutes from home, I find myself wishing there was a “Set for destination” feature enabled in my cyclist’s setting on the Driver app. My next ping could quite easily propel me further away still, and then I’d be looking at an hour-long bike ride back; time and physical effort for which I’m not paid. As Bloodworth writes in Hired about his experience of driving for Uber, it is, all in all, “a peculiar sort of freedom”.

A cyclist wearing Deliveroo’s teal-and-white outfit swooshes past me and I feel a pang of impostor’s guilt when he gives me a hard-set, comradely nod. I’m emblazoned with UberEATS’s neon green, but there’s no company loyalty in this world. Why would there be? We’re not technically employees. My gear, which consists of a branded thermal-box backpack and a waterproof handlebar mount for my smartphone, was offered to me after my background check was cleared. Since I’m not classed as an employee, the equipment isn’t enforced as official livery. I would have been fully within my rights to buy or rent an unbranded box backpack, but found that UberEATS provided an irresistibly convenient deal for acquiring the necessary gear: a bag, insulated box backpack, phone holder and phone charger would be sent to me immediately, if I agreed that the £120 deposit would be deducted in-app from my pay in weekly £20 instalments once I started riding. Having only been UberEATSing sporadically, I find the balance currently displayed under the “Earnings” tab in my app shows a negative sum.

This deposit notwithstanding, Uber does have past form in exploitative lending. In 2015, it set up the Delaware-based subsidiary Xchange Leasing, a car rental division offering subprime leases to Uber drivers who were cleared to drive but had such poor credit ratings that they couldn’t buy or lease a new car (Uber demands that drivers’ vehicles are no more than eight years old).3 The conditions of these leases were well within loan-shark territory: after a $250 down payment, drivers were required to pay up to $200 in weekly instalments for 36 months. “The best part?” said a promotional video: “Payments are automatically deducted from your Uber earnings.” At the end of the three-year lease, Uber would release drivers from the agreement but retain the $250 deposit. Should the driver then wish to buy the car in question, they would need “to fork over the residual value of the car, which could run many thousands of dollars,” journalists Eric Newcomer and Olivia Zaleski explained in a 2016 report for Bloomberg. (Maybe “the worst part”?) One driver in their investigation reported that the $183 weekly payments he made for his Xchange-leased 2015 Toyota Prius alone would amount to $28,798 by the end of the three years, while the automotive research company Kelley Blue Book values the fair purchase price of a 2015 Prius at $21,985. (Nope, found a worse part.) From mid-2015 and onwards, once hundreds of thousands of drivers were already locked into leases, Uber began slashing base rates, forcing drivers to work longer hours in order to keep up their weekly auto-deducted payments. (Bloody hell, a worse part still?)

In the autumn of 2017, Uber began to dismantle its car-lending services and in January this year, it announced that it had sold off Xchange Leasing. Despite skinning drivers, it turns out that the two-year‑old subsidiary was wildly unprofitable, losing an average of $9,000 per vehicle.4 Dara Khosrowshahi, who took over as CEO of Uber last summer following Kalanick’s resignation, is currently attempting to “shift gears”, in the words of a chummy 2018 New Yorker profile, and transform Uber from a not-so-nice band of “brilliant jerks” (as a play by Joseph Charlton describes Uber) into a nice “traditional company”. Shutting down Xchange Leasing, I assume, was part of this change-of-gears, as was the aforementioned “180 Days of Change” campaign. In Europe, Uber has just announced it will partner with the insurer AXA to provide accident cover and parental pay. The new Driver app, too, was announced with something of a personal apology from Khosrowshahi: “Drivers are the heart of our service. But along the way, we lost sight of that. We focused too much on growth and not enough on the people who made that growth possible.” The new app, Khosrowshahi promises, is one measure among many that Uber is taking to “improve the driving experience”.

I’ve arrived at my destination and consult the app on my phone. I want to find the address details and check whether my eater has left any specific instructions but am interrupted by a fresh ping: “Meal, 8 min away.” I have yet to complete my current job but the Driver app has a forward-dispatch algorithm activated by default. This means the app sends through new opportunities in much the same way that Netflix and YouTube encourage viewers to stay on their platforms by automatically previewing and counting down to new content at the end of current episodes and clips. The latter, known as “post play”, is widely believed to encourage binge watching. Many drivers and cyclists enjoy forward dispatch, as it cuts down idling time and increases earnings by providing back-to-back jobs. I find it just a little stressful, though, and, because I’d initially brought up the app to access my eater’s details, I begin to fumble. The “accept” field occupies roughly a third of my screen (in contrast to the much smaller “No thanks” button in the upper left-hand corner), so I accidentally take on “Meal, 8 min away” when I try to swipe the ping off. No, no, cancel! My thumb furiously flits across the screen. This will affect my cancellation rate, which is displayed on my profile. In the meantime my current eater, who is able to track my whereabouts on their interface, must be wondering what I’m doing standing on their doorstep for several minutes. I decide to cancel, only to be taken to a multiple-choice screen asking me to explain why. None of the options say, “You’ve purposely designed the app so that it’s easier to accept than to reject a ping,” so I tick “Other.” Another screen. Argh. “Please specify.” With trembling fingers, I type out “Accident”. I’m back at the page with the eater’s details.

Forward dispatch will remain a key feature in the new Driver app. In fact, it will be augmented, explains Zack Gottlieb, a member of Uber’s design team working on the project: “We’ve just made it easier to see when something’s coming up. It’s part of the whole redesign experience[…] it’s just more apparent.” Notifications will also be presented and stored in an inbox of sorts, rather than shown fleetingly in the feed that takes up much of the landing screen. “Earnings is one of the top things that drivers care about, but this information [is] spread all over the place in the [current] app,” says Bryant Jow, the latest app’s lead designer. “In the new app, we’ve put earnings front and centre. At the top, [there will be] this earnings tracker that updates after every trip.” I’ve no doubt that the decluttering and clarification of the app’s features will improve the user experience, but these changes also point to the increasing gamification of the interface, a development that has been going on for some time.

In an in-depth report for the The New York Times last year, Noam Scheiber accounted for the ways in which, with the help of “hundreds of social scientists and data scientists, Uber has experimented with video game techniques, graphics and non-cash rewards of little value that can prod drivers into working longer and harder”. Forward dispatch was one of the algorithms Scheiber discussed, alongside other psychological inducements, such as suggesting earnings goals that drivers have not set themselves. “You’re $10 away from making $330 in net earnings. Are you sure you want to go offline?” a typical push notification could read, Scheiber reported. “Go offline” and “Keep driving” are the options given, with the latter already highlighted. “Some of the most addictive games ever made, like the 1980s and 90s hit Tetris, rely on a feeling of progress toward a goal that is always just beyond the player’s grasp,” Scheiber explained. It’s a phenomenon that Natasha Schüll, the anthropologist and author of Addiction by Design (2014), has called the “ludic loop”. Uber’s argument, as laid out by Amodeo in the The New York Times’s report, is that “any driver can stop work literally at the tap of a button – the decision whether or not to drive is 100 per cent theirs.” But in light of Uber’s heavy investment in gaming techniques and psychological incitements, one could argue this is a little bit like telling an internet addict just to log off Facebook, or a gambler to leave the casino. Eli Solomon, described in the article as a “veteran Uber and Lyft driver”, said: “The whole thing is like a video game.” I wonder what he’ll make of the new Driver app, with its Super Mario-like coin counter.

The eater comes downstairs to meet me (“we strongly recommend that you do not go into the eater’s home,” the instruction video has warned), carefully avoiding eye contact as they take their food – an interaction I’ve become accustomed to when UberEATSing. After they disappear, I bring up the app, swipe to confirm the delivery, and formally end the job. I take a swig of water while the app is calculating the fare and check my profile to see if the eater has left an in-app star rating, compliment or badge for me. Uber’s feedback system allows for written complaints and issue reports from end-users, as well as entirely non-essential badges signifying compliments such as “Excellent Service” (represented by a sparkling diamond) or “Great Route Choice” (a red pin on a map). In his report, Scheiber interviewed a driver whose earnings working with Uber were so low he had to apply regularly for payday loans. Still, when asked about the badges he’d accrued, the driver “practically gushed”: “I’ve got currently 12 excellent-service and nine great-conversation badges.”

I have no badges. My rating is a solid five stars, though, and this is because all profiles start out with a full five-star rating. Fall below 4.5 and you risk being deactivated. The app’s feedback system, then, not only adds to the gamification of the work, but also plays a managerial role. “Passengers are empowered to act as middle managers over drivers, whose ratings directly impact their employment eligibility,” note Rosenblat and Stark in their paper. “This redistribution of managerial oversight and power away from formalized middle management and toward consumers is part of a broader trend in flexible labor.” The term that’s increasingly being used by academics to describe this phenomenon is “algorithmic management” – a setup that owes much to pre-New Deal Taylorist management models, the chief difference being that “the app is both the factory and the boss”, as journalist Alexis Madrigal summarises in The Atlantic.

“We’re not setting the price, the market is setting the price, and we have algorithms that determine what that market is,” Kalanick told Wired in 2013, when asked about Uber’s surge-pricing tactics. It’s a quote that attempts to pass off the platform as a neutral conduit of existing market forces, and it is only within such an understanding of algorithmic management that drivers and couriers can be considered to be independent agents. When one takes into account what Rosenblat and Stark call the “information asymmetry” present in the technology, however, the picture becomes much more complex. This asymmetry is tilted in Uber’s favour, from blind driving to the perennial fluctuation of base fares, which drivers must accept if they want to keep on using the app. It affects the end-user too. Consider the landing screen in Uber’s passenger app, where a number of cars will typically be shown circling your location. In 2015, an Uber Help staffer explained to a passenger that this shouldn’t be taken as a literal representation of cars in your area. “I know this seems misleading to you,” they wrote, “but it is meant as more of a visual effect more than an accurate location of drivers in the area. It would be better of you to think of this as a screen saver on a computer.” Yet, this is the page which, viewed at a glance, will often determine whether a passenger decides to book a car or not.

On the driver side, glitches in the app appear to nudge them to go on for longer: “For the past 4-5 days,” writes one on, “my last ride button [which temporarily pauses forward dispatch] has been playing hide ’n’ seek by blinking off and making it hard to engage.” Another driver comments: “Yes, Uber’s shady tactic to keep you online. Simply toggle the online on/off button and you’re good to go.” This exchange recalls an account from an earlier era of unfettered capitalist enterprise, James Myles’s 1850 book Chapters in the Life of a Dundee Factory Boy: “In reality there were no regular hours, [as] masters and managers did with it as they liked. The clocks at the factories were often put forward in the morning and back at night, and instead of being instruments for the measurement of time, they were used as cloaks for cheaters and oppression. Though this was known amongst the hands, all were afraid to speak, and a workman then was afraid to carry a watch, as it was no uncommon event to dismiss anyone who presumed too much about the science of horology.” A technology, be it a clock or an app, is not neutral if the tools that allow for its manipulation lie entirely in the hands of one party.

The app has calculated my pay for the job I’ve just completed: a little over £5. From ping to completion, it’s taken me just under half an hour. Ten pounds per hour is not the sort of earnings I’d expect for someone with the word “partner” in their job title. But across the on-demand economy, such euphemistic nomenclature is the norm. Cyclists and drivers on Uber’s platform are called delivery and driver “partners”; the low-wage pickers in Amazon’s warehouses are “associates”; warehouses are not warehouses but “fulfilment centres”. And when you’re dismissed, reports Bloodworth, who spent a few months working as a picker in Amazon’s Rugeley warehouse, you’re not sacked or fired but “released”. “Almost everything that had a name,” he writes, “was given a euphemism.”

Once you get past the Orwellian overtones of this language, it becomes clear that it serves a very specific purpose. Just as algorithmic management collapses a host of functions previously held by human beings into a single interface, this nomenclature folds the structurally antagonistic notions of worker, supervisor and boss into one undifferentiated soup. “Jeff Bezos [Amazon’s CEO] is an associate and so are all of you,” Bloodworth recalls being told on his first day in Rugeley. When I ask Wilkinson about this type of language in the on-demand economy, she describes it as “smoke and mirrors”: “It’s key to remember that the only legal tests around the employment status relationship are to define ‘employee’ and ‘worker’. The reality is open to interpretation. ‘Self-employed’ status exists, but it’s got no statutory definition.”

I decide to call it a day, and start pedalling home. I’ve not made much money, but neither, it would appear, has Uber. The strangeness of the whole setup is that Uber, despite being valued at a staggering $72bn in February this year, is not a profitable company. In fact, it’s nowhere near breaking even: a loss of $670m was reported in 2014, $1.5bn in 2015, $2.8bn in 2016, and $4.5bn last year. It has, however, raised more money from private investors before going public than any other technology firm, and this lets it subsidise its competitive low fares. Such is the brave new logic of late-capitalist enterprise: expand aggressively by burning through stupefying amounts of investors’ cash and then, once competitors have been driven out of business, begin to hike up fares. Uber’s backers aren’t worried: “The trend is good,” said the political consultant and Uber investor Bradley Tusk last year, perhaps mindful that Bezos’s Amazon was unprofitable for its first two decades. The picture looks a little different for the drivers and couriers providing services through such platforms as Uber, however. As demonstrated by Deliveroo’s recent decision to grant £10m worth of company shares to its employees, but nothing to its self-employed fleet of cycle couriers, much of what is “good” about “the trend” hinges on employment status. A sleek new app will go some way towards improving the driving experience for Uber’s partners, but unless some of the interface’s information asymmetries are levelled, the gesture will remain cosmetic.

1 In this world, your customer is not your customer but your eater. I will return to the significance of this nomenclature.
2 Passengers get 10 to 15 per cent off the price and Uber takes a 35 per cent cut rather than the usual 25.
3 Yes, subprime loans. No lessons learned from 2008, then.
4 Lesson from 2008 learned.