One fiercely hot evening in February, Sukiman rode his motorbike along the Trans-Sumatra Highway, a dusty strip of road that runs north-south the entire length of the island, part of the Indonesian archipelago. Sukiman, a 20-something Gojek driver, who asked Rest of World not to use his real name, had picked up a passenger in a nearby town to take her to Amplas, the main bus terminal in the city of Medan. It was a routine trip but also a significant one.
Drivers on the service earn points by successfully completing a certain number of journeys. In many Indonesian cities, one trip on Gojek GoRide, a ride-hailing service, earns one point, while a food delivery run earns one and a half points. Drivers told Rest of World that if they “score” 12 points in a day, the company bumps the driver’s daily earnings to $7.70 (110,000 rupiah), regardless of how much they’ve billed for the day. That means a driver who makes 12 trips worth $3.50 (50,000 rupiah) can more than double his earnings; while one who makes 11 trips at the same total value gets no bonus. This was one of the last trips Sukiman needed to hit 12 points that day.
When he arrived at the destination and dropped the passenger off, Sukiman’s phone beeped. “Manipulative order detected.” Gojek’s system had decided that the journey wasn’t real and that Sukiman had booked it himself in order to get more points.
“I escorted a real person, but the system said it was a fake transaction,” he told Rest of World. He tried to contact the customer service of Gojek, but without evidence, he couldn’t prove the trip was real. Gojek then suspended Sukiman’s account for three days as a punishment.
Millions of Gojek drivers like Sukiman rely on the company’s algorithm to put food on the table every day. Lured by its flexible working hours and relatively high pay when the company first launched, many left their work in the informal sector or in manufacturing jobs to join the gig economy. Their labor helped to turn the company into one of Indonesia’s first “decacorns” — private companies worth more than $10 billion. Gojek is now on the cusp of a merger with another tech giant, Tokopedia, a move that analysts expect will lead to the combined entity listing on a major stock exchange.
Underlying the company’s success is an automated management system that is supposed to drive productivity within its “partners.” Every time a Gojek driver completes an order, they receive a fee from the passenger and points from Gojek. Passengers evaluate the driver using a star rating. Without a good rating, the driver doesn’t get as many orders. They also get punished for violating a long list of company rules, such as creating fake orders and asking more money from passengers. Drivers who break the rules have their accounts suspended. Repeat offenders are banned from the app. But alongside these automated management processes is a system of incentives and punishments that drivers have to compete with in order to earn their pay. The way that these rewards are allocated are opaque to the drivers, unhitching their pay from straightforward metrics of money earned for hours worked.
“The company controls and intensifies labor production by imposing reward and punishment,” says Aulia Nastiti, a researcher in political science at Northwestern University, in Illinois. “It is all monitored by technology.”
In its response to Rest of World, Gojek compiled a long list of the support it offers to its driver partners, including insurance, scholarships for drivers’ families, and basic health kits for protection during the pandemic. The decacorn also said that it obeys local laws and regulations in its interactions and contracts with drivers. It also presented studies pointing to the company’s positive contributions to Indonesia’s economy and drivers’ prosperity.
Gojek drivers, whose livelihoods depend on their income from the app, liken working for the algorithm to playing a kind of game, albeit one that’s far from entertaining. Rest of World spoke to a dozen Gojek contractors across Indonesia, who said that they are increasingly being driven to meet ever more challenging targets and to play by rules that can be changed without consultation or warning. This, one driver based in Surabaya, told Rest of World, makes it more like gambling. “The dealer always wins,” he said. “And the players always lose.”
When Joko, another Gojek worker who requested anonymity, rode up to a supermarket on the outskirts of Medan on another scorching hot February day in the city, he found a line of motorcycles already waiting. The branded green jackets and helmets left on the bikes indicated that several of his fellow Gojek riders had got there ahead of him. The riders often congregate together to eat and socialize while they wait for a job. Joko crossed the forecourt to wait by the main gate. There, he found Sukiman, whose account suspension had now been lifted. They waited together for another two hours, but as the time passed, nothing popped up on their screens.
Waiting is what most Gojek drivers do most of the time. The previous day, Joko had only picked up one job, worth 56 cents (8,000 rupiah). “This is the impact of pandemic: the number of customers are declining and more people have become Gojek drivers,” he said. “What can I do?”
That much is easy to understand: the pandemic has left millions of Indonesians out of work and led many more to sign up for gig work, meaning that more drivers are competing for fewer passengers — although the slump in demand has been mitigated by surging use of the platform for food delivery.
But contractors and researchers told Rest of World that the reality that some drivers fail to get any work at all, while others seem to thrive, could have more to do with Gojek’s algorithm than with dwindling demand.
In a 2019 podcast produced by Gojek, company leaders, including co-founder Nadiem Makarim, argued that the algorithms used by the company to decide what jobs go to what drivers are neutral and objective and that the way the app allocates labor to contractors is fair. Raditya Wibowo, then Gojek’s head of product development, told listeners that he often heard complaints from drivers with no orders. Most of the time, “they weren’t actually facing any technical issue,” Wibowo said. Rather, the system was penalizing them for bad behavior. Either the drivers were just in the wrong place or “repeat offenders,” Hans Patuwo, then the company’s chief operating officer, added.
But the algorithm that decides what jobs go to which workers, and which drivers get downrated and penalized, is only as good as the data that goes into it. Last year, Jessica Basukie, who was then a researcher at the London School of Economics and Political Science, published a study into how Gojek’s algorithms worked. Among other concerns, she found cases where drivers had been suspended after they were called “ugly” and given a bad rating. Rest of World found similar examples in conversations with drivers and in their private social media posts. Gojek provides a system for drivers to appeal, but to cancel a suspension, the customer needs to withdraw their complaint. A suspension can take a driver off the road for several days, cutting their income.
“This also signifies a power asymmetry, in that the company prioritises the customers, and drivers do not have the power to defend themselves before they get suspended,” Basukie wrote.
Drivers also said that the system doesn’t take into account basic factors that would be standard for a human manager, such as who has been waiting longest for a ride. “In the past, drivers with the closest location to customers got matched. But it doesn’t work like that anymore,” says De Lasblang — not his real name — a Gojek driver from a small city in East Java. Gojek monitors how many orders the drivers have completed, passenger ratings, and how long they spend online. De Lasblang performs above average: 98.8% performance, five stars from passengers, and 225 online hours. “But still, that doesn’t influence the number of orders coming to me,” he says. “Only Gojek knows this mystery.”
This can make it hard to hit the daily point threshold that triggers a payout from Gojek. “It is very difficult to reach 12 points nowadays,” says De Lasblang. In the days before he spoke to Rest of World, he said his daily income had ranged from $1.12–$3.50 (16,000–50,000 rupiah), before paying for fuel, data, and food. Most days he hadn’t reached the 12-point target.
Drivers have developed their own language to describe the inequities in the system, using words like gacor (an abbreviation of “gampang cari orderan,” or “easy to get orders”), to refer to accounts who easily get rides, and gagu (Indonesian word for “muted”), for accounts who rarely get any.
Tae Wan Kim, an associate professor at Carnegie Mellon University in Pittsburgh who has studied algorithmic management and gamification, told Rest of World that this phenomenon represents “algorithm unfairness,” a serious problem in the ride-hailing industry. “Companies like Uber say their algorithms are determined purely by supply and demand, and there is no room for manipulation. But no one knows [if] that’s really true,” Kim told Rest of World.
No study has investigated why some Gojek drivers are gagu and some are gacor, but a mathematical simulation done by Anikó Hannàk at the University of Zurich found that, when there are more drivers working for a ride-hailing company in a city, the matchmaking algorithm naturally widens drivers’ income inequality. “Our most surprising result is that drivers with the same qualifications and working hours can end up with vastly different incomes,” she wrote.
Unsurprisingly, some drivers have tried to cheat the system so that they can trigger the 12-point payout.
“It’s a natural thing in a game,” said Arfive Gandhi, a researcher at the Faculty of Computer Science at the University of Indonesia. “People do whatever they can to win. Although I admit that it’s morally wrong for drivers to cheat.”
For Gojek drivers, that could mean creating a fake order using their own phones, giving them another trip and another point toward the payout. One customer in the Greater Jakarta area, who asked to remain anonymous, told Rest of World that a driver had asked her to book a second, false trip after he had dropped her off. “He asked me to pay in cash so that Gojek wouldn’t be able to detect [the deception]. In reality, I didn’t give him the money,” she said. The passenger agreed to book the trip because she felt sorry for the driver.
Gojek also has systems to detect these kinds of tricks and automatically decides whether a trip is fraudulent. Sukiman’s experience shows that it is not foolproof.
“Algorithms are trained by datasets from human behaviors,” Kim said. “Thus, they are not perfect.”
The pandemic has been hard for many people in Indonesia. In mid-2020, the economy went into recession for the first time in more than two decades. An estimated 3 million people have lost their jobs. Gojek, however, has been boosted by the sudden shift of large swathes of the economy online.
In November 2020, the company announced that all of its main products were now generating positive margins, putting profitability within reach. The company has spent a decade and millions of dollars on subsidies and discounts to build its market share and create a vast user base across Southeast Asia. It has also paid out bonuses and incentives to drivers to entice them from rival services.
As the company transitions to profitability, it’s squeezing contractors. Gojek has already cut bonuses. Drivers used to get additional incentives for hitting 14 points in a day, but, starting in around 2017, those were progressively cut, until they were ended entirely in 2020.
When the pandemic hit, Gojek applied a new “berkat” (favor) system, with which the company determines a maximum income for drivers. In February, all drivers in North Sumatra had to collect 12 points to get $7.70 (110,000 rupiah). But in March, Gojek classified drivers into four different levels — basic, silver, gold, and platinum — based on their last month’s rating, points, and performance. They also re-based the points system, multiplying it by 100 “to make it easier for partners to calculate their additional point schemes in the future.”
Many drivers think this is unfair. Drivers in the basic category now get their income topped up to $5 (72,000 rupiah) if they collect 900 points. However, drivers said that the math doesn’t make sense. On average, it might take only six trips to earn $5, which wouldn’t be enough points to trigger a payout. That makes the bonus threshold essentially meaningless. “Do they intend to give berkat or are they just joking with us?” one driver said.
Gojek has the latitude to experiment like this because there is a willing supply of drivers, displaced from their old jobs by the economic crisis. But the company has also created a game that drivers are hooked on, one in which it can change the rules to suit itself.
Nastiti says she doesn’t deny that there are positive aspects to Gojek’s model. “The technology is not 100% grim,” she said. But algorithmic management is disruptive, and it’s proliferating fast. Its impacts need to be better understood, so that ethical conflicts can be managed. “Our main concern now is how we are going to adapt [to this technology], so that no one is harmed.”