Twenty Years Into the Fake Digital Economy, Internet Companies Depend More on Offline Business — and Carry More of Its Costs

It used to be "clicks and mortar." Now the mortar keeps getting heavier — letting tech companies remake traditional industries, but also saddling them with the mortar's costs.

Uber's IPO turned into a disaster.

Over the past five years, only 10% of tech companies that went public in the US closed below their offer price on day one. Uber fell 7.2% on day one and another 10.37% on day two, its market cap plunging from a planned $100 billion before the listing to $62.2 billion — already below its 2015 valuation.

A drop like that can't be explained away by bad timing. Of the nearly 3,500 companies on the Nasdaq, fewer than 3% fell harder than Uber over those two days.

As the most valuable private tech company in the world, Uber's disastrous debut became a signal.

"This is a signal. Uber, once the most sought-after company, had to confront reality, and that tells you something." Tim O'Reilly, who has been active in tech circles since the 1990s and helped popularize the concepts of open source and Web 2.0, told The Information that Silicon Valley used to assume that if a company was worth that much, it must be doing something exceptionally right — but the two things don't always go together.

This isn't just Uber's problem. Lyft, the world's third-largest ride-hailing company, which went public in the US a month earlier, has already lost a third of its market value. Didi, the second-largest ride-hailing company by valuation after Uber, hasn't gone public yet, but it's mired in the same losses — more than 10 billion yuan last year.

Uber and Didi aren't just the most valuable startups in the US and China; the ride-hailing business they're in has long been held up as a representative of the so-called digital economy (or "new economy").

The convenience and impact of ride-hailing can't be denied and shouldn't be. The very existence of Uber and Didi demonstrates how internet technology can remake a traditional industry — before them, taxi companies were typically confined to a single city, never reaching a market of hundreds of millions or more than a billion people.

But the trouble both companies now face reveals exactly what kind of thing ride-hailing is: a fake digital economy. It looks like it lives inside an app, seemingly sidestepping the operating costs traditional taxi companies bear by not employing drivers directly — but in reality, the money spent running things offline never actually goes away.

Meanwhile, Uber and Didi have told their founding stories using the growth and prospects of the earliest pure internet companies, only to end up burning cash to subsidize enormous offline expenses just to sustain internet-style growth.

Similar fake digital economies are everywhere among the new business models born over the past decade — group buying, O2O, bike-sharing — each one more dependent on offline operations than the last, and each one's problems surfacing earlier than expected.

This history of the fake digital economy's development is, in fact, the history of internet companies penetrating industry after industry. As Marc Andreessen put it, "software is eating the world" — but after software eats the world, the real-world spending doesn't disappear.

Ride-hailing is a model case of the internet remaking a traditional industry — and a representative of the fake digital economy

Ride-hailing companies generally fall into two types. One is the platform model, like Didi and Uber, which pools idle vehicles and drivers to match rides with passengers and earns commissions on each transaction. The other is the self-operated model, like UCAR (神州专车) and Geely's Caocao Mobility, which build, buy, or lease their own vehicles and assemble dedicated driver teams to serve passengers directly.

Pooling idle resources is clearly easier to scale than building your own fleet. According to Aurora Mobile's ride-hailing market report, Didi had 66 million monthly active users at the end of 2018, while UCAR and Caocao combined had less than 10% of that. Uber's monthly active users reached 91 million by the end of 2018.

But the bigger the scale, the bigger the losses. Didi lost 10.9 billion yuan in 2018, more than four times its 2017 loss. Uber's revenue for 2016 through 2018 was $3.845 billion, $7.932 billion, and $11.27 billion respectively, with losses of $370 million, $2.2 billion, and $1.85 billion in those same years.

The main reason for the losses: the subsidies to drivers and riders never stop.

Last year, Didi's commission rate was 19%, but 7 percentage points of that went right back out as driver subsidies — 11.3 billion yuan paid to drivers over the year. Another 10 points went to other operating costs, and 4 points to taxes and payment fees, adding up to 21% — against which it still lost 2%.

Since going public, Uber hasn't disclosed the specific cost of acquiring and subsidizing drivers, nor the churn rate among drivers and riders (which determines whether Uber can grow healthily without subsidies). Withholding that data says something in itself.

On Monday morning US time, as the stock kept sliding and employees watched their holdings shrink, Uber CEO Dara Khosrowshahi wrote in an email to staff: "Remember how tough things were for Facebook and Amazon after their IPOs. Look at them now. We'll follow the same path."

As it has before, Uber compared itself to earlier successful internet companies. Amazon and Taobao built online retail platforms; Facebook and Google built ad-distribution platforms on internet services; Tencent distributed games over the internet. These companies built their business models entirely on digital channels and reaped rich rewards — that was the start of the digital economy.

Whether it's Alibaba and Tencent, which have made money for years, or Amazon, which went unprofitable for a long stretch, their most important shared trait is that capital kept flowing in net, not out. That gave them bargaining power over suppliers and partners up and down the chain — they could grow their business using other people's money, further widening their competitive edge.

Amazon is the textbook case. Even in its unprofitable years, it kept generating growing net cash inflows through sales, which let it steadily raise operating efficiency and pour cash into the infrastructure future growth would need — automated warehouses, cloud computing R&D — building a moat.

So for many self-styled "platform" companies, the real problem isn't whether they'll eventually turn a profit. It's that even after building scale advantages and competitiveness, they still can't stabilize — or even pin down — their costs, expenses, or liabilities.

Uber and Didi, running ride-hailing, don't have to pay for buying or leasing cars, but they can't avoid subsidizing drivers and riders. That subsidy money disappears the moment it's spent — nothing like Amazon plowing early cash into infrastructure that would raise efficiency and generate returns down the line. And ride-hailing platforms typically settle up with drivers weekly; they can't hold onto suppliers' money for months the way e-commerce does — Amazon has managed to hold merchants' funds for an average of more than two months, using that float to invest in new businesses and technology.

By comparison, the cost of pooling idle drivers turns out to be much higher than expected. As companies like Didi and Uber build up scale, cultivating a stable driver supply requires those cars and drivers on the platform to become steadily more professionalized and steadily more dependent on the platform. And to stand out against rivals doing the same thing, the platform has to be cheaper too — which means it has to keep subsidizing drivers just to hold onto them.

In the end, it was UCAR — the company that built its own driver team and bore the high cost of a self-owned fleet — that started showing signs of profitability: a loss of 3.67 billion yuan in 2016, a loss of 300 million yuan in 2017, and a profit of 270 million yuan in 2018.

Once a platform holds onto its drivers, it still has to deal with management and safety issues. So what once looked like an external partner gradually becomes an internal cost of the company, and the lawsuits or claims users might bring effectively become the company's liabilities too.

And even holding onto drivers doesn't guarantee holding onto riders. From January to April 2018, Meituan briefly offered riders bigger subsidies than Didi.

The result: in Shanghai, within less than 20 hours of launching taxi and express-car services, Meituan's ride-hailing arm announced daily orders had topped 100,000; by the third day, daily completed orders topped 300,000, and its Shanghai market share reached a third. Now Amap has emerged as another competitor siphoning off traffic to other ride-hailing platforms.

The same competitive dynamic plays out in the US — however large Uber's market share, it still faces Lyft's challenge. By the same token, Uber's early push into global markets hasn't stopped one competitor after another from carving out share in China, Russia, and Southeast Asia.

This has to do with the nature of the product. Most people choosing a ride-hailing app only care about whether they can get a car quickly right now — not whether the same app will work the next time they travel abroad. As long as a new competitor can achieve enough density in one city, it doesn't need to worry about the network effects an industry giant spent tens of billions building globally over the years.

That means the ride-hailing business built on pooling outside drivers is a market with very low customer loyalty — offer subsidies and low prices, and you can grab market share. The scale advantage an incumbent leader supposedly holds turns out to be fragile; it takes continuous capital burn just to maintain the appearance of an edge, and the industry's development amounts to an endless cycle of price wars and consolidation.

This is also the biggest difference between Didi and Uber on one side and Amazon on the other. Amazon can tie up capital from suppliers up and down its chain, adjusting payment terms to invest money owed to suppliers and merchants into new technology and new businesses first. Didi and Uber, by contrast, can neither delay payments to drivers nor predict when a new rival will suddenly launch a subsidy war. So the competitive edge their scale is supposed to deliver turns out to be very thin — they need constant fundraising and subsidies just to hold onto that fleeting advantage, while their own net cash flow never really accumulates.

The costs shift, but they don't disappear — this is common across every type of new-economy company

Struggling with cash flow while racing ahead by ignoring costs it should be bearing — bike-sharing fits that description too.

Bike-sharing only appeared in 2016. Mobike, a pioneer of the industry, rolled out bikes in Shanghai in April 2016 and hit the streets of Beijing in August, kicking off the industry's fever-pitch wave of "raise funds — build bikes — deploy bikes." But it had no idea how to turn a profit, and it overlooked the costs that urban management, road traffic, and public safety could impose.

Bike-sharing's business model is easy enough to understand: build the bikes yourself or have a manufacturer build them, fit them with smart locks, put them on the street, let users scan a QR code with their phone to unlock a bike, and charge them once they're done riding — 0.5 or 1 yuan per half hour.

But smart locks can be expensive to make. Yong'an Xingxing (永安行), a listed company also in the bike-sharing business, disclosed detailed manufacturing costs in 2017: the lock and controller accounted for 50% of a bike's manufacturing cost, while the frame accounted for just 15%.

Bikes were expensive to build and cheap to ride — which raised questions about profitability from the start.

Even so, though the business model itself wasn't especially novel, this was 2016 — China's internet demographic dividend hadn't run out yet, and the whole society was still in the grip of a startup and investment boom. Any new project with an internet angle drew capital eagerly.

Scan-to-ride bikes were packaged as a new wave of the internet remaking urban mobility. In the four months after Mobike entered Beijing, it landed four rounds of funding, ranging from tens of millions to over a hundred million dollars. Peers like ofo, Bluegogo, and Youon also kept raising money and rapidly deployed bikes across cities.

Bike-sharing companies rushed to flood multiple cities with bikes without factoring in the cost of occupying public space. Local governments gradually realized the impact all those parked bikes had on road traffic and public safety, and began impounding bikes deployed in violation of rules, requiring designated parking zones, and even banning companies from deploying more bikes, switching to a quota-and-bidding system instead.

Even setting aside the urban management costs that were bound to come sooner or later, bike-sharing's business model had a fundamental flaw: bikes depreciated faster than expected, and average daily usage time was shorter than expected.

Shanghai's current bike-sharing standard requires bikes to be retired after 3 years. Manufacturing costs for Mobike and ofo bikes have been reported at various figures; taking the lower estimates — 550 yuan per bike for Mobike, 270 yuan for ofo — as a baseline for calculation:

Retiring bikes after three years means annual depreciation of 183 yuan for Mobike and 90 yuan for ofo. Mobike's CEO previously disclosed that bike utilization was below 10%, meaning each bike was used roughly two hours a day. Excluding rainy and snowy days, and assuming an average of 300 usable days a year, four rides a day at 0.5 yuan each, one bike would generate roughly 600 yuan in revenue a year.

But that's before even accounting for enormous marketing and operations costs — like the tricycles you see hauling bikes around the streets — and bike-sharing also has a high loss rate. According to an ofo employee, ofo's loss rate was around 20%, Mobike's around 10% — meaning 10% to 20% of bikes simply vanished every year. Add in the subsidy wars bike-sharing companies waged against each other to grab users, and for a long stretch they barely charged anything at all. It was only natural that a business that looked profitable on paper turned into a bottomless money pit in practice.

To this day, bike-sharing — a business with low capital spending that designs its own products and even builds its own manufacturing capacity — has had to pay the price for a deeply flawed business model once the frenzy died down, on top of the costs imposed by urban management and road traffic. The industry is now a mess: ofo users still have no idea when they'll get their deposits back, and Mobike, after being acquired by Meituan, racked up 4.5 billion yuan in losses between April and December last year alone. As for the second-tier bike-sharing companies, nearly all of them have already collapsed.

Online education is another sector that was supposed to save money but ended up burning even more.

In theory, moving students' extracurricular tutoring and adult vocational education online could break the constraints of time and space, diversify teaching methods, make "teaching to the individual" possible, and let institutions save on rent and some classroom costs.

But in practice, marketing and student-acquisition costs stayed stubbornly high, students lacked discipline, and teaching quality became harder to guarantee.

51Talk, Liulishuo, Hujiang, Ambow Education, and Puxin Education — internet education companies across the board have kept posting losses. Only New Oriental Online turned a profit, riding on the "New Oriental" brand — itself the product of years of offline development.

In China, uneven and scarce teaching resources, combined with schools' habit of uniform instruction, have kept demand for extracurricular tutoring alive, giving rise to a massive offline tutoring industry.

But because course content varies by region and for other reasons, the industry remains highly fragmented — even New Oriental and TAL Education, the two nationwide players, hold less than 5% market share between them. Offline education's constraints on students' and teachers' time and location have capped how fast it can grow — and that's exactly the starting point online education promised to improve on. But the new problems online education brought gradually canceled out the costs it was supposed to save.

The first problem is teaching quality.

Many platforms have built real-time scoring and assessment systems. Liulishuo, for instance, scores users' spoken pronunciation in-app and evaluates their skill level after every lesson, letting users track their progress in real time and using AI to correct their spoken English. Hujiang's website offers a dedicated courseware system: as students answer quiz questions or click interactive elements to trigger a lesson, the system uses their responses to build a customized learning plan.

The systems can be sophisticated, but without a physical classroom to hold students accountable, the lesson ends the moment they close the page. On top of that, most platforms randomly match students with teachers for live or recorded classes, leaving teachers with little understanding of each student's situation. That's clearly not teaching to the individual.

Basic doubts about teaching quality have made parents slow to embrace this new format, forcing online education platforms to spend far more on marketing to win over customers — driving student-acquisition costs sky high.

Whether it's Hujiang, with more than 8 million paying users, or Liulishuo, which markets itself as teaching English with AI, marketing expenses have exceeded total revenue, and losses have kept widening: Hujiang lost 280 million yuan, 422 million yuan, and 537 million yuan in 2015, 2016, and 2017 respectively; Liulishuo lost 89 million yuan, 243 million yuan, and 488 million yuan in 2016, 2017, and 2018.

It's not just pure online education companies losing money, either — even training companies that used to turn a profit offline started losing money after moving online. Ambow Education, which once had a near-monopoly on adult diploma education, shut down all of its offline campuses — built up over nearly a decade — in 2014 to pivot to live online classes, and its marketing costs then shot up. That shift shows up in its headcount: Ambow had more than 4,000 sales staff, six times the size of its teaching staff, driving enormous sales and administrative expenses. From 2015 to 2017, Ambow's marketing expenses were 333 million yuan, 504 million yuan, and 1.352 billion yuan, with losses of 318 million yuan, 254 million yuan, and 919 million yuan.

"Clicks and mortar" was already being talked about 20 years ago — take Ctrip, though back then it was only marketing that stayed offline

The issue Ctrip chairman James Liang cares most about today is probably demographics. But 20 years ago, when he founded Ctrip, what he was thinking about was how to bring Silicon Valley's "clicks and mortar" model to the company.

Ctrip was founded in 1999, on the eve of the dot-com crash. In 2003, it listed on Nasdaq on the strength of 29.19 million yuan in net profit over the prior nine months, and its 89% first-day gain set a record for the exchange going back to 2000.

According to a piece James Liang wrote for the Chinese edition of Forbes in 2013, the key to turning a profit and going public quickly in the early 2000s — when internet companies were dying fast — was "call center plus internet."

"Clicks and mortar" was really just another way of describing running online and offline operations together. Under this model, a company combines its offline (mortar) and online (clicks) channels, or goods and services. Offline could mean a paper catalog, or it could mean a call center that took orders by phone.

Ctrip back then had a 500-person call center team providing bilingual service around the clock. Their job was to sell customers the product they needed within 180 seconds. To make that possible, Ctrip rented 30 phone lines, each with 8 extensions, letting it handle 240 calls at once. The company also hired 40 in-house telesales staff, whose job was to call prospective customers to pitch products and services, tell occasional customers about the company's latest offerings, and encourage them to use its services more often.

This approach was directly tied to the makeup of China's internet and mobile-phone users at the time. In 2003, China had about 68 million internet users but more than 250 million mobile phone users.

In the first nine months of 2003, Ctrip took in more than twice as many orders by phone — placing calls and taking them — as it did online.

After going public, Ctrip stopped disclosing the cost of its phone-based labor. But the share of total revenue spent on service costs, mostly the call center operation, fell from 30.2% in 2000 to 13% in 2003, then climbed back to 17% in 2005.

Unlike the ride-hailing companies that came later, Ctrip bore the cost of traditional phone sales, but that was only on the marketing side — the business processes behind it, whether scraping airline and hotel prices or processing orders online, were thoroughly internet-native.

From street promotion to delivery fleets: internet companies have grown more "fake digital," but that's also what let them truly break into an industry

Marketing offline the way Ctrip did went on for a long time — in the early days of online gaming companies promoting themselves in internet cafés, and later during the early street-promotion pushes of group buying and the taxi-hailing wars.

The "hundred group-buying sites war" around 2010 best captures how seriously internet companies collectively took offline work. Four or five hundred group-buying sites opened branch offices across the country, moving local restaurants and lifestyle services online. In less than a year, the number of participating group-buying sites doubled to four or five thousand.

Because group-buying sites and the products they sold were highly homogeneous, and competitors were plentiful while the number of merchants and consumers' purchasing power had a ceiling, what determined who survived the fight was how much a site subsidized merchants and consumers, and how efficient its street-promotion team was.

Meituan was famous for its street-promotion team. It insisted on doing "self-operated" street promotion — visiting merchants one by one, running just one deal a day but guaranteeing quality, and funneling traffic to one merchant at a time.

Gan Jiawei, Meituan's then-head of operations, later said publicly that many group-buying companies subsidized restaurants and movie tickets to rack up gross transaction volume, but the moment the subsidies stopped, the volume disappeared with them. According to his account, Meituan didn't just subsidize consumers — it also approached top merchants in various niches, promised them a sales figure, but demanded exclusivity in return, then paid out of pocket to hit that promised number.

Gan Jiawei worked at Alibaba early in his career. He and Didi's Cheng Wei were both once members of Alibaba's street-promotion team, the "China Supplier Iron Army." "China Suppliers" referred to Chinese manufacturers; Alibaba tried to convince Chinese manufacturing companies to set up shop on its cross-border trade platform, and its street-promotion team had to both persuade these traditional manufacturers to start selling online and verify they weren't scammers.

A salesperson who worked on Alibaba's street-promotion team for a year and a half told QDaily that after joining in early 2010, he was sent to Alibaba's Hangzhou headquarters for a month of paid training, mostly memorizing Alibaba's corporate values and getting familiar with the product.

Over the phone, he still remembered a few of Alibaba's slogans from back then, like "do simple things repeatedly, do repetitive things persistently."

Once he officially started, he was assigned to Alibaba's Songjiang operations post in Shanghai. He clocked in at eight and headed out to visit clients by eight-thirty. "On the way, we were usually on the phone lining up the next day's visits," he said. "Each day, we'd typically visit one pre-scheduled client in the afternoon, then do 'cold visits' to nearby manufacturers." Cold visits meant showing up unannounced.

Quite a few people on his team chartered cars to visit clients, at 200 yuan a day. By his estimate, a salesperson had to make over a thousand calls a day to have a shot at hitting the quota: persuading two companies a day to sit down for a meeting.

Later, whether it was Alipay or other internet companies, promoting a product never skipped the street-promotion step. Didi and Kuaidi also sent people out to stake out street corners, holding up QR codes for passersby to scan and download the app, then explaining how to get free subsidized rides. At the same time, they sent people to places where taxi drivers gathered to walk them through installing the app and opening a bank account, step by step.

Uber, founded back in 2010, went so far as to buy iPhones for drivers early on to replace their feature phones.

Didi initially just let taxi drivers take orders through its app, but taxi drivers were still controlled by taxi companies, and those companies were monopolized in every city by a handful of local players — never going to bring Didi reasonable revenue.

Similarly, Meituan-Dianping's early work on group buying and organizing offline restaurant information could already turn a profit, but it was just one of many marketing channels for a restaurant — it didn't truly determine how many orders a restaurant got. At its biggest, that industry topped out at roughly the combined scale of Yelp and Groupon in the US. Today those two companies have a combined market cap of $4.5 billion — less than a tenth of Meituan's.

Eventually, Didi followed Uber's lead and started dealing with drivers directly, taking on enormous operating costs — but because it could then decide whether a driver got orders and what kind, it could finally take a cut of every ride.

Meituan, meanwhile, started running delivery itself, taking on the cost of organizing delivery logistics. Meituan's prospectus shows that in 2015, when it started doing food delivery, it lost 123.7 yuan for every 100 yuan it took in from users. But precisely because of that massive loss, Meituan gained significant power over how many orders a restaurant received — restaurants now had to pay commissions and also advertise on the delivery platform.

Once these platforms stopped stopping at the marketing stage offline — that is, stopped just doing street promotion — and paid the price in massive losses, they became less and less "digital economy." But that's exactly how they penetrated the key operational core of an industry, building companies in ride-hailing and food delivery on a scale that once would have been unimaginable.

There's nothing inherently wrong with a "fake" digital economy — the problem is that internet companies keep assuming a purely digital economy's low costs for their future

"Digital economy" was always a fuzzy concept, and so is "fake digital economy." Whether a company is run well, and how much value it creates, isn't decided by whether it's "real" or "fake."

Alibaba, once cautious about offline investment and mocking JD.com for building its own logistics network, now has courier-industry investments that rank among the top three by market cap in the whole express-delivery sector. And after a string of failed online fresh-food ventures, Hema has been opening stores one after another inside shopping malls.

Likewise, Amazon has been making its e-commerce business more and more platform-like to lower its own costs, while after years of failing at online fresh groceries, it spent more than $10 billion acquiring Whole Foods and even started opening offline bookstores.

These offline investments aren't happening because these companies don't know they'll burn cash — they're happening because this is the cost of truly breaking into certain industries.

The problem is that as internet companies increasingly use offline operations to build competitive advantages, they keep assuming that because they're internet companies, the necessary costs will somehow disappear.

Under that assumption, Uber decided to sustain the kind of breakneck growth expected of an internet company, aiming for the scale a low-marginal-cost company could achieve. After all, from founding to IPO, it kept sketching a picture of a global mobility market worth $5.7 trillion (equivalent to 7% of global GDP) — a startup's dream where grabbing just a small slice would be enough.

To sustain that growth, Uber set a relatively low take rate per ride and kept pushing down the price of every trip — which made drivers easy to lose. So it kept subsidizing drivers to reduce churn, and subsidizing riders to fight off competitors.

Bill Gurley, one of Uber's early investors, publicly warned back in 2016 about the danger of tech companies chasing growth at any cost, arguing that so-called unicorns should accept they weren't worth as much as claimed, cut their valuations and costs, and go public sooner rather than later. Uber didn't do that. In the end, both its CEO and Gurley faded out of the company's operations after a series of conflicts. Having burned cash for three more years, Uber missed a better market window and became one of the worst IPOs in the history of internet companies.

The problems Uber faces today are widespread among this generation of internet companies pushing into the real economy — and most of them fall far short of Uber even in efficiency, or in the value they deliver to investors and consumers.

As software eats the world, "clicks and mortar" has become standard equipment for internet companies. And internet companies can't avoid taking on more and more of the mortar's costs.

Cover image via: Pixabay, Laith Abdulkareem on Unsplash