Today, after a tortuously long wait for DHL to deliver my gear to my new “Hinterhaus” apartment (their first two attempts failed), I finally worked my first session for Deliveroo in Berlin. This session, from 11:30 to 13:00 in Neukölln, was the only one still available when I was granted access to Deliveroo’s booking system at 17:00 last Monday. Because I had to cancel a few sessions I had previously booked (when I still expected my gear to arrive on time) less than 24 hours in advance, my late cancellation rate had already taken a hit before I even completed a session, which meant that I was now last in line when it came to booking sessions for the following two weeks. Those with the best “statistics” (i.e. attendance rate and late cancellation rate) get priority access at 13:00, followed by the second tier of riders who get to book their shifts at 15:00. By the time riders like myself get access to the booking system, I’ve noticed that very few to no sessions are left, depending on the zone you check into. So much for the idea of working whenever you want… Sure, if I would want to bike all the way to Wedding or Grunewald there may still be some availability, but since I don’t get paid to bike those kilometers I’m sticking to zones closer to my apartment in Kreuzberg: GAK (Gendarmenmarkt, Alexanderplatz, Kreuzberg), Neukölln, Schöneberg, and perhaps Friedrichshain – which is supposedly a very popular and busy area during dinner time. Moreover, I have noticed that you actually do get quite a few last minute session offers when you request to be notified when a spot opens up in the zone and time-frame you prefer.
A little while ago I received such a notification, offering me another hour of work in Neukölln at the start of the dinner peak later today, which I gladly accepted. I also decided to opt in for the trial that Deliveroo is currently running in Germany, between 15 November and 20 December: the introduction of distance-based fees in combination with the end of “blind dispatch” (which hides the customer address from the rider until s/he picks the food up at the restaurant) and an overview of the full fee – which, importantly, will now change with each delivery. I had received an email about this trial about a week ago and already heard/read some riders discuss it, usually in a positive light. So far, riders seem to generally like the prospective of receiving more money when they had to travel longer distances to complete an order, which only seems fair, and they are happy to get more information about an order before they decide whether or not to accept it. Better pay and more transparency, who could object to this?
The fact that short-distance orders can now also pay less than the previous standard payout of 5 euros per order is also taken into consideration, but I am not yet getting the sense that many riders here are worried about this possibility. This could partly be explained by Deliveroo’s pledge that it will compensate riders who can show that they have made less money than they would have made with the “old fee per delivery rate” for the number of orders they complete. Until 20 December, Deliveroo “will pay you the difference”. But what will happen after the trial ends? Looking at the UK, which is usually one step ahead of other Deliveroo markets, one notices that distance-based fees are no longer an opt-in experiment in its trial phase and has been introduced as the new status quo applicable to all riders. As perhaps could be expected, there is no longer any mention of compensation for earnings that are lower than they would have been in the previous system. Such compensation is ostensibly not necessary, given that “in reality, you will see higher fee (sic) for the majority of orders, and far higher fees for longer distances.” For those who happen to get mostly short-distance orders, a minimum delivery fee of £3,60 currently forms a guarantee at the bottom end of the dynamic earnings spectrum.
In the “reality” assumed by Deliveroo, this minimum delivery fee as well as the fee per kilometer (which is not disclosed and apparently varies per city) will remain stable. This, however, is not the reality I have witnessed and learned about in New York City (my previous fieldwork site) and, from what I’ve seen on Facebook, neither is it the reality experienced by Deliveroo riders in the UK. In NYC, couriers (who don’t identify as “riders”) with years of experience told me how they have seen their earnings drop over time, no matter which app they used. In a city where variable, distance-based payouts have been the market norm for at least 3 years, food delivery platforms have repeatedly reduced both the minimum delivery fee and the fee per mile, according to couriers I spoke with. (There is no official documentation by the company itself, which shows such fee reductions, but couriers have been pursuing their own documentation using screenshots, shared in closed Facebook groups.) Meanwhile, although Deliveroo’s distance-based fee system was only recently introduced in the UK, riders there are already sharing their concerns online now that they are seeing slight drops in their minimum delivery fee. The question on these riders’ minds is what to expect over the coming months: will these fees be further reduced or will there be an upwards readjustment? Their peers in NYC may already have the answer to that question, but I am not sure how well it will be received.
NYC couriers have generally coped with decreasing payouts in the best way they can: work more hours and work smarter. Working smarter in this case means being attentive to available bonuses (Uber calls them “Promotions” while Caviar uses the term “Milestones”), which can significantly compensate for the reduced base fees/rates and can even make couriers quite a good sum of money – if they manage to meet the applicable terms and conditions. It is here that we get to the main argument of this research note: as the regime of stable (hourly or piece-rate) wages gives way to dynamic earnings schemes, and as these dynamic earnings tend to drop over time, gamified incentives will become increasingly important to 1) keep workers from leaving the platform, 2) nudge them to work on specific times and in particular areas, and 3) motivate them to take on more work (i.e. work longer and faster). In Europe, the design and implementation of incentive schemes is still in its relative infancy, judging by the current operations of food delivery platforms like Deliveroo. But now that the company has introduced its new payment regime, I predict that it will soon start to develop and test new incentives, which will not only become increasingly central to how it algorithmically manages its workforce but are also likely to become more personalized and adaptive, based on the growing body of data it collects from each rider. I will return to this last point below.
So far, Deliveroo has been offering a limited set of bonus payouts in Berlin (each market has its own variations on the company’s general incentive scheme), which apply to all riders equally as long as they are working a session and area subject to a bonus. For instance, on Friday, Saturday, and/or Sunday, riders get 50 euros extra when completing 50 orders and receive a 120-euro bonus after doing 100 orders. On Sunday, which is usually the most high-demand day of the week, doing at least 10 orders will get you 15 euros extra, and it also happens that riders receive push notifications telling them that they can earn between 1 and 2,50 euro extra on each order they complete that day – during a particular timeframe and in a specific zone. It should again be noted here that these so-called “fee boosts” are universal, in the sense that all riders receive the same bonus offers, regardless of their performance data or “statistics”. They are also static, in the sense that the boost per order is fixed during the active period. Such incentive characteristics are likely to change over the coming time and indeed they already are changing in the UK market, as demonstrated by a text message received by a Deliveroo rider there: “To help you earn great fees when you ride with us we’re introducing a new type of fee boost that gets bigger when you deliver more orders!” In addition to adaptive fee boosts based on rider performance, a look at the NYC market suggest that such incentives will become progressively more customized.
In New York, couriers working for Uber Eats regularly check their app to inspect their portfolio of weekly Promotions, which are likely tailored to the specific data profile of each rider – potentially containing information on their work history including order acceptance patterns, number of completed orders (in the last week/month), average completion time, and login patterns. I write “likely” and “potentially” here, as Uber does not disclose how it determines or allocates Promotions and couriers have no idea what makes them eligible for particular offers. What they do know, however, is that other couriers who operate in the same area during the same times do not get the same Promotions, which is occasionally a source of frustration in the Facebook groups I frequent. Personalized incentives generate a greater sense of inequality and competition among couriers, while they at the same time also excite and entice a workforce whose job can often be tedious, tough, and poorly remunerated. The gamification of this job, through incentives such as “earn $15 extra by completing 5 trips” or “1.8x – 4.0x Boost per trip”, then helps to alleviate such tedium and stimulates couriers to keep pushing themselves in order to reach their income goals (see Sarah Mason’s recent article in the Guardian for a fascinating account of platform labor gamification).
Such goals can grow increasingly ambitious, even while they may not always be more lucrative when considered in terms of unit economics. This is especially the case when the incentive consists of an income guarantee rather than a bonus payment. Take for instance the following recent offer from Postmates: “Earn $800 for your next 100 deliveries, guaranteed. Ends in 5 days.” Not only does this come down to $8 per delivery, which is not particularly high in the NYC market, incentives like this often come with terms and conditions that stipulate “the rules of the game” in each particular case – demanding, for example, that a courier has to accept x% of the orders being offered. Even when such stipulation is not effective, however, this courier is likely to accept most of the orders that come his/her way because s/he has to complete 20 orders each day (if s/he is even able to spread them out over the full five days). This will include much longer trips than s/he would otherwise choose to accept, which may even have to be completed by taking a subway ride (as NYC couriers are sometimes known to do). In other words, s/he will have to work his/her ass off while chasing such an incentive, while only being guaranteed $8 per order. Still, the prospect receiving $800 after five days of hard work, guaranteed, is something many couriers do not want to pass up on; at least it offers some income security in an industry notorious for feeding off contingent and precarious labor. It is therefore all the more painful when the goal is eventually not reached because a courier came up just a few orders short, as I’ve been told happens quite frequently. In the gig economy, a guarantee is not always guaranteed.
What also tends to occur, according to some NYC couriers I’m still in touch with, is that incentives become less appealing and lucrative over time. On Facebook, a courier responded to Postmates’ offer by pointing out that the number of orders that needed to be completed to receive the $800 guarantee used to be 80 instead of 100, to which another courier replied that at one point it was only 60 orders. To them, Postmates was just trying to get people to work more for less money, which turned them off. Less experienced couriers, however, this same guarantee may seem like a good deal. And this is where we get to the final point I’d like to make here, which also returns me to Berlin and to Deliveroo’s current experiment: food delivery platforms (like other labor platforms) are increasingly leveraging the behavioral data they collect about their “partners” to figure out what, for each of them individually, constitutes a “good deal” at specific times and under particular circumstances. If a courier accepts the incentive that offers $800 after delivering 100 orders, will s/he accept 110 orders for the same amount or will the guarantee have to be raised accordingly – with what amount and when? How many orders is s/he willing to complete for $500, or for $1000, and what should be the timeframe and conditions? If s/he ignores the incentives that s/he is offered, how do they need to be adjusted for him/her to start pursuing them? With some regularity, NYC-based couriers share stories of getting better Promotions not just when the weather is bad but also when they haven’t been logged on for a while. They feel like Uber is trying to win them back. On the other hand, some couriers notice that they tend to get worse Promotions after they have pursued a high number of them during the previous weeks. In this case, they wonder if Uber is trying them out somehow.
While many of these tendencies are based on personal experiences and speculation, so far lacking systematic documentation, there is nevertheless reason to believe that food delivery platforms will increasingly use behavioral data to create incentives that are more responsive to the changing profiles and proclivities of individual couriers, to the extent that such agile incentives will more successfully perform the three functions mentioned above. This, then, explains why Deliveroo in Berlin (and elsewhere) has decided to provide its riders “with the full delivery journey”, made “available before you accept the delivery.” More order transparency also means more data on riders’ situated decision-making processes: what will it take for a rider to accept this type of order?; how far is a rider willing to go for this payout – and what if we add this particular incentive? Upfront dynamic pricing is already well established in the ride-hailing business (where both sides of the two-sided market ask “am I willing to take this ride for this price?”) as well as in other industries, and now Deliveroo is introducing it to food delivery in Berlin (following lessons learned in the UK).
I already had my first taste of this new reality when I logged in for a second session later in the day, having opted in to the trial after first session ended. I first accepted an order that I could now see was relatively close but did pay less than 5 euros. Then I received a couple of orders which paid between 6 and 7,50 euros but which would take me much longer to complete, with one order offering me 7,37 euros to deliver the food to an address in Mitte, more than 4 kilometers from the restaurant (which itself was another kilometer away). Needless to say I rejected that one. I actually ended up rejecting quite a few incoming orders, after which I each time had to select one option from a list of pre-defined reasons for rejecting. Meanwhile, I accepted a number of easier orders that allowed me to stay in my area, but these all made me less than the former standard of 5 euros. I also couldn’t help but think about what this behavior said about me, as a rider: what kind of conclusions would Deliveroo (its algorithms, its data scientists and perhaps also its operations managers) possibly be able to draw from the data input I was now constantly generating? Were my decisions going to influence the next orders I was going to receive? Or was I just being paranoid? There was no way of knowing yet, but I will surely be keeping an eye out, taking notes and screenshots. And come to think of it, I may also take Deliveroo up on that compensation deal.