AI may solve Meituan’s offline headaches

The party is over for China’s second-generation tech giants. Fueled by easy money, new markets, and lower transaction friction, they have fought for their market share, burning money as they went. Now it’s time to pay the piper. While economists are still undecided about the exact figures, it is clear that the Chinese economy isn’t doing very well. In 2018, according to official data, China’s GDP (gross domestic product, a measure of the market value of goods and services in a country) grew by a sluggish 6.6%, the lowest since 1990.

When first encountering the Chinese tech ecosystem, many people are surprised by the scale and speed. Amazed by the work ethic, pragmatism, and ambition, their attention is drawn away from the risks that such scale and speed entail. Too much, too fast has been the downfall of many a Chinese tech entrepreneur. From Ofo to LeEco, China’s tech is littered with the bodies of the fallen, both big and small. China tech entrepreneurs, as Kaifu Lee has put it, are best compared to gladiators: locked in a life-or-death battle for survival. Growing up in a scarce but rapidly developing environment, they’ve learned not only to move fast and be aggressive, but also to build their moats by any means necessary.

At TechNode, we’ve written quite a bit about the “2VC” model, the many restructurings, and the existential challenges that established players are currently facing. There’s a lot going on and I recommend you read those pieces to get a full understanding.

Meituan is experiencing its own contraction pains, but like many who came up in the mobile and O2O (online-to-offline) revolution, it’s the offline component that is causing it the most headaches. However, it’s the latest in technology that could solve most of their growing pains.

Meituan’s moat

CEO Wang Xing is perhaps the most representative example of China’s “fake it until you make it” copycat culture. All of his companies (except perhaps ticketing platform Maoyan) were directly copied from already established Valley darlings, including Facebook and Twitter. It wasn’t until Groupon took off that Wang finally found a model that worked. Backed by Tencent, Meituan was the only survivor of the 2010’s group-buying boom-bust cycle. Groupon’s Achilles heel, however, turned out to be Meituan’s greatest strength.

In the pre-mobile internet era, merchant information online was extremely unreliable. Many Chinese friends would ask me, naive American that I was, why I trusted the information on websites. If I wanted to actually get reliable information, I had to talk with a real human on the phone (quite difficult, given my Chinese language ability at the time). Nowadays, Dianping may be a rich repository of merchant information, but back then—long before it was bought by Meituan in 2015—it was still merely a platform for user-generated reviews. Investors in the Valley and in China saw group-buying as a chance to finally collect all that “last mile” data about local businesses. However, the difference between the markets and business models ultimately came down to one of China’s greatest discoveries: monetizing a network through shopping. By leveraging an aggressive but relatively cheap workforce, they were able to lock in merchants, offer competitive discounts, and continually improve the benefit to both consumers and businesses. Since then, it has become the “Alibaba” of O2O, connecting users with a plethora of services ranging from food delivery, travel, entertainment, car maintenance, and even furniture. Now, you can also access cab and private car rides as well as some of the best bike rentals in China.

The deliveryman in the room

O2O and the “sharing” economy have been great for Chinese consumers and workers. Not only have consumers gotten increased convenience at very bearable prices, but the boom has also encouraged small business growth in the form of contractors and novel business models as well as providing well-paid jobs in the service sector. However, people just aren’t scalable or sustainable in the same way as software.

In order to build their moats, the two delivery giants, Meituan and Ele.me, have spent incredible amounts of money. Last year, Ele.me promised to spend up to $400 million to increase its market share. As of June 2018, Meituan Dianping reported in its IPO prospectus that they spent almost RMB 7 billion on sales and marketing, around 54% more than in the same period in 2017. As of the third quarter of 2018, the last financial report Meituan Dianping has published, it spent a little under RMB 5 billion on the same, putting it on track to exceed the 2017 total of almost RMB 11 billion. According to the IPO prospectus, sales and marketing costs were mostly attributed to user incentives, promotion, advertising, and employee benefits expenses tied to sales and marketing staff involved in expanding its delivery network. According to the prospectus, it costs around RMB 7 per order to pay the delivery driver. However, as large cities like Beijing and Shanghai aim to curb migration, that labor force will soon start to dry up, creating a lot of room for doubt around the cost of labor in the future.

Under the hood

O2O champions Didi and Meituan are experiencing similar problems with scalability, and given China’s AI boom, they also have a similar solution. Both companies have invested significant amounts in order to apply the latest technology to their scaling problem and it seems to be working. While overall costs are high, Meituan reports sales and marketing expenses as declining as a percentage of revenue “driven by improving economies of scale.” Much of this derives from the application of AI to its logistics, efficiently matching drivers with orders. I’ve seen this firsthand. I once placed two separate orders for merchants located not far from each other. Meituan’s system automatically assigned the same delivery person for both orders and I still got them within the promised time. Didi, for their part, made a lot of noise in early 2018 about their “AI Brain,” which was designed to use its big data stores to help the government solve traffic problems. Given its larger challenges, it’s not clear how many problems it has actually solved to date, but I can say anecdotally that its driver assignment and navigation systems have improved dramatically.

Didi, Meituan Dianping, Ele.me, and the myriad other O2O niche services will probably never be known as AI companies, at least not like Baidu or Bytedance. However, these are indisputably the companies at the forefront of applying AI to real operational problems. As the Chinese economy slows, this approach may not be a choice for them. Their dominance, and their very survival, depends on whether they can create effective AI algorithms to optimize how they use the increasingly expensive and scarce offline resources.