images-2On Tuesday night, I participated in a discussion panel at the University of Maryland on current and future innovations in Supply Chain Management (SCM). While most of the topics we discussed were positive developments, I did warn that some SCM advances have aggressive characteristics, by which I mean that they have the capacity to harm certain human populations that they affect. I used the example of retail workforce scheduling software and the havoc it can wreck in the lives of many retail workers. I also suggested that Uber might fall into this category, since my hunch was that the current Uber model also places extraordinary strains on its workforce. Apparently, researchers have already begun to study these strains, as I found out today when I read a new paper by Alex Rosenblat, of the Data and Society Research Institute, and Luke Stark of NYU.

Their paper, Uber’s Drivers: Information Asymmetries and Control in Dynamic Work, is a fascinating examination of the impact that Uber’s analytics and algorithms have on the work life of Uber drivers. The paper outlines how the Uber platform is not just a demand-supply matching system but also a sophisticated system “of surveillance and control over workers’ behavior.” The authors explain that this system is fundamental to Uber’s ability to structure what they call “indirect control” over its workers using information asymmetries over critical work aspects such as rider demand, rider reviews, and spot pricing dynamics. Moreover, the authors claim that Uber relies “heavily on the evolving rhetoric of the algorithm to justify these information asymmetries to drivers, riders, as well as regulators and outlets of public opinion.”

The major conclusion of the paper is that Uber’s platform has created a workplace that gives “freedom and flexibility” to drivers in exchange for working in a regime of “surveillance and resistance,” the inner workings of which are kept hidden from Uber’s drivers. As the authors note: “In the Uber system, the labor drivers do is actually shaped by two primary factors: the employer’s use of surveillant practices to effect ‘soft control’ (Deleuze, 1990; Boltanski & Chiapello, 2007) over otherwise flexible independent contractors, and corresponding practices of resistance developed by those workers in the system (Ball, 2010; Levy, 2014).” The heart of the Uber system is, of course, its proprietary platform. This platform has three important design principles, and understanding each principle is key to grasping the impact of the authors’ conclusions about what it’s like to work in the Uber environment.

The Uber labor platform design principles:

  1. Constantly monitor drivers, even when they are not working
  2. Create information asymmetries between Uber and its drivers
  3. Replace middle managers with (anonymized) customers

Critical to design principle number one is to ensure that Uber’s drivers’ are flowing data to the platform at all times. As the authors note: “Uber drivers continue to generate useful data for Uber even when they are not carrying a fare (known as “dead miles”) because they relay data back to the central platform from which inferences can be drawn about traffic patterns, and which feed into supply and demand algorithmic analyses.” This “automatic production” of data, the paper notes, “even when they are not being paid, marks the on-demand platform-economy’s departure from a traditional service economy” where “off-the-clock” time created no value for an employer. The authors go on to note that the constant digital connectivity “of platform-based work enables a type of continuous, soft surveillance by employers/platforms. It also enables more precise, efficient matching between ‘supply and demand’ in real-time by the platform/ employer with a broader view of a multifaceted system, while simultaneously maintaining a socio-legal distance between an employer and workers.” In these  words, the brilliance of Uber’s first design principal begins to appear: Uber is able to get more and more data/value from its drivers while at the same time building an algorithmic platform/firewall between its drivers and itself.



Uber Surge Pricing Screen (Image Source: Paper)

Principle number two is well-illustrated when one looks at how surge pricing works for Uber drivers. The paper succinctly illustrates the problems with this part of Uber’s model:

Surge pricing, however, is unreliable: notably, pricing is based on what a passenger sees on screen in their location, not a driver’s position. Drivers travel to surge pricing zones in search of fares advertised at a given rate, but they can and do receive ride requests from passengers in other, adjacent areas. A driver may enter a zone that is surging at 3.5x, but receive ride requests at a lower surge rate, such as 1.5 based on the passenger’s (not the driver’s) location. In forums and in interviews, some drivers describe this as a type of wage theft: they are advertised one rate of pay through heat maps, but given another. Others offer the company rhetoric, which is that surge pricing is subject to dynamic change and that the rate they see for their area may not reflect the rate at which passengers request them. Some drivers report that passengers are gaming the system by placing their pick-up location pin outside a surge zone, and then calling drivers to redirect them to their actual pick-up location. Drivers also noted that they would sometimes converge en masse at a surging area, and find that supply was no longer too low — the surge would disappear. Some drivers reported experimenting with trying to game these algorithms, and many developed responses to surge pricing based on their experience with its duration, reliability, and potential reward in their respective locations. As various drivers become familiar with the features and functions of the app, they have begun to advise each other and to ask about surge; “don’t chase the surge,” is offered in forums as guiding advice to new drivers.

This model results in a world where drivers take on the risk of possibly phantom surges and passengers — basically, Uber’s business risk — without any of the rewards of being an Uber manager or owner. It is, in effect, an elegant use of information asymmetry to enable a risk transfer mechanism from Uber to its drivers, all bathed in the language of driver/partner and entrepreneurship. “This discourse of entrepreneurship,” the paper goes on to note, “in the tech sector is the legacy of a Silicon Valley environment where highly skilled and mobile workers could take on risks and trade-offs to be part of the startup world (Neff, 2012, p. 24), but this rhetoric of risk has effectively been retooled to suit a contingent of lower-income workers who are recruited to perform service labor, not highly-skilled technical work.”

Uber Driver Rating Summary (Source: Paper)

Uber Driver Rating Summary (Source: Paper)

The last principle noted above is best illustrated by Uber’s famous driver rating system. The system allows passengers to rate drivers on a scale from one to five stars.  However, most passengers probably do not realize that (a) these ratings directly determine driver access to the platform and (b) even a slight deviation from a five-star score is considered poor performance. As the authors explain:

Passengers are not made cognizant of the fact that a 4.6 represents a hazardously low rating for a driver, and subsequently it always appears to a passenger that all the available drivers are good performers [Emphasis mine]. Some drivers directly nudge passengers to prompt good ratings such as by adding five star stickers to a visible place near their windshield. This behavior is partial compensation for Uber’s overt lack of communication with passengers as to the value and instrumental character of driver ratings.

Rosenblat and Stark go on to make the interesting point that Uber is trying to recreate the “seamlessness” of the app experience in its management of drivers by basing driver rewards (access to demand) and punishments (getting kicked off the platform) not on worker-boss interaction but on anonymous feedback filtered through the platform’s algorithms.  “Yet the disintermediation that the app enables between the different actors who use the Uber system,” note the authors, “facilitates channels of communication that are ripe for information mismanagement: indirect management gives Uber power over its workers for which it is often not held accountable.” Thus, Uber has replaced the traditional labor command and control function — usually known as “middle management” — with an algorithm that is an amalgamation of customer likes and dislikes. In other words, it is the customers who command and control driver behavior and not a traditional management layer. This ingenious design principle not only lowers Uber’s operating costs but allows the company to anonymize middle management in a way that must make a lot of other companies quite envious.

At this point, it’s interesting to ask if Uber’s pioneering algorithm-as-boss model is unique to this company or could be used in many other contexts? I think the answer to that question is a resounding yes. Uber’s developers have designed not just a brilliant ride-sharing platform but the foundation of an entirely new management paradigm. Moreover, the platform’s three core principles — collect all data possible, design in and exploit information asymmetries, and outsource labor management to platform/customers — could easily be applied in countless other settings — from restaurants to hospitals to schools.

As I finished the article, I thought back to when my first consulting firm moved from traditional, in-person, partner-consultant performance reviews to a data-driven and anonymous web-based system. The system generated so many complaints that a meeting was held where the managing partner kept telling us that it was “the process” that had generated some erroneous bad reviews, that “the process” may have resulted in some questionable promotions, and that “the process” was probably to blame for some unfair bonus allocations. I remember walking up to him after the meeting and saying: “You know, when I started at this firm I wanted to be a partner, but now I think just want to be The Process.” I’m sure Uber’s drivers would say the same thing about their algorithm/boss: invisible, all-powerful, and the harbinger of a brave new world indeed.


Read more:


Posted by Carlos Alvarenga

Carlos Alvarenga is the Executive Director of World 50 ThinkLabs and an Adjunct Professor at the University of Maryland's Smith School of Business.


  1. Coauthor’s last name is Stark, with a ‘k’ 🙂


    1. Thank you for the head’s up. I have corrected the spelling here and on the version on LinkedIn. Cheers, Carlos


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s