ByteDance is obsessed with daily active users and revenue growth and has developed a management style around “listening to the data.” But in a market where value is being redefined, it is now a victim of its own methods.
By XIAO Fang
In a year where layoffs and cost-cutting are the talks of big tech, ByteDance seems a remarkable exception. Job bulletins are flooded with ByteDance posts. Marketers and data engineers bomb their contact lists with “Join Our Team!” emails.
To be straight, there are layoffs at ByteDance. In June, 3,000 people were let go in education-related businesses alone. Gaming and Operations are faring hardly better. What sets ByteDance apart is that hiring is more aggressive, and firing more ruthless. Rumor has it that the average employee stays for only seven months, which is largely confirmed by those working there.
The practice of hiring more people than needed is a proud tradition that ByteDance makes no secret of. It’s not uncommon to have three or four teams working on the same project and vying for credit. Competition sparks innovation, as managers like to remind their underlings. But in reality, competition also begets corporate infighting and waste.
The company makes new and ongoing projects visible to everyone and allows open bidding regardless of function or title. Theoretically, this “breaks silos.” But it also alerts everyone that someone else is eyeing their turf.
More often than not, it takes a ton of time, back-and-forth emails to form a team and even more emotional damage to claim credit when the projects are pulled off. As ByteDance matures and previous breakthroughs become business-as-usual, many so-called “innovations” are nothing more than incremental improvements that don’t merit such pain.
During a Q&A session with Douyin CEO ZHANG Nan last year, an employee raised the question. Social Media and Content did essentially the same thing, he asked, isn’t it neijuan (involution, self-defeating, unproductive competition) and a waste of everyone’s time and effort?
Zhang’s answer was a categorical no. It’s not unproductive, he said, because daily active user (DAU) numbers have increased. So keep up the excellent work.
The obsession with DAU is coded in ByteDance’s genes. Like many tech companies, ByteDance evaluates performance using the OKR (Objectives and Key Results) framework. But here, goals are set and assessed results every other month instead of every quarter, and progress is measured in highly quantifiable terms, DAU and revenue, in particular.
A former Meituan manager spent eight months researching viability without any new product materializing. This is extreme but by no means uncommon in big tech. But at ByteDance, the only way to know whether an idea works is to get your hands dirty, or to AB test as they say in tech jargon. If DAU or revenue is higher in the experimental group (Scenario A) than in the control group (Scenario B), it is a good idea.
Douyin (TikTok) and Toutiao, apps that went more than viral and arguably transformed the internet, were products of such a purely data-driven approach. Don’t ask why or how something works. User behaviors are messy, irrational and counterintuitive. If data – excuse me, DAU and revenue – says it works, then it does.
If an idea is greenlit, it is adopted at breakneck speed, with no small amount of ferocity and cruelty. Whenever TikTok entered a new country, ByteDance would advertise at sometimes four times its rivals' intensity. In the Philippines, it outspent Kuaishou to irrelevance in just three months.
The formula worked miracle after miracle, leaving ByteDance more obsessed with data and speed than ever. By the time it ventured into education in 2018, there was little doubt that the fight would be quick and easy.
In a little more than a year, 10,000 people were toiling to bring in more DAU and revenue for 20 new education apps, and ByteDance wanted to hire 10,000 more. Don't worry about over-hiring. Just fire people if they don't bring in results.
This applied from first-year analysts all the way up to executives. GuaGuaLong (AI teaching kindergarteners), Qingbei Wangxiao (human teachers teaching K12 classes) and Xiaoma AI (AI teachers teaching K12 classes) each went through numerous department heads who failed their bi-monthly OKR reviews.
The problem was that education products, by nature, need more time to accumulate users and get them hooked than, say, goofy videos. Tech news site Late Post once reported that the team behind Aixue (AI-generated customized problem sets) never had enough time to sign up enough schools for AB tests before their following OKR reviews were due.
"Senior managers didn’t know what they were doing," a former employee said, adding that education products require a lot of know-how, but whoever understood the industry and saw what ByteDance had been doing wouldn't get themselves into such a mess.
Just like that, ByteDance education went through cycles of mass hiring and mass firing without outcompeting anyone before the K9 tutoring school crackdown last year put everything on pause.
Founder Zhang Yiming once boasted that he didn’t and wouldn’t hold onto any given approach. But the truth is that ByteDance is so obsessed with data that it refuses to let it go even when it becomes clear that not all objectives or critical results can be quantified.
When executives at Xigua (long videos) set out to “invigorate the content creator community,” their underlings were at a loss. How do you define a vigorous community? How do you measure “vigor” in numbers? What about beauty? Or the effectiveness of storytelling? How do you quantify them?
The same questions are also asked in gaming, instant messaging (Lark) and, basically, every industry where ByteDance hopes to replicate Douyin and Toutiao’s success. Without being given clear answers, employees have learned to pick OKR-friendly tasks and leave the slow, hard, potentially open-ended but nevertheless necessary ones unattended. Department heads who don’t get to pick easy tasks are simply removed. Xigua alone has had at least three CEOs since 2020.
Corporate culture also deteriorated. Managers were too busy protecting their own turf to mentor young workers. Young workers spend a significant proportion of their days meticulously documenting every minor decision just in case of finger-pointing and backstabbing. Everyone is burned out inventing quantifiable and achievable objectives and worrying about not achieving them. The office is still full at 9 p.m. every day, but many workers say they never have enough time to act and learn new things.
After Douyin and Toutiao, ByteDance has never created anything as viral. Granted, there was regulatory backlash and bad luck, but doubts have crept in that ByteDance, obsessed with short-term deliverables, is unfit for long-term planning and strategy execution. Or it is simply too complacent with its success to rethink objectives, resulting in a more saturated market and less tolerant regulatory environment? In the new regime where value is measured not only in DAU and revenue, no one can afford to be lazy, not even ByteDance.
