Baidu Hit a Ceiling Over a Medical Scandal, and This Time Its Usual Patchwork Fixes Won't Work | Top 15 Annual Report

A company that only ever chased short-term returns, unable to bear the responsibility its scale demands.

One week before the 2017 Lunar New Year, Baidu announced the biggest management shake-up in its 17-year history.

Lu Qi would become Group Chief Operating Officer, taking full charge of technology, product, operations, marketing, and sales across the company. The role was new, and every business-unit head would now report to him directly.

For Baidu this was a seismic change. It made Lu Qi the company's effective operational chief going forward.

And Baidu badly needed a seismic change.

In the third quarter of 2016, Baidu's advertising business shrank for the first time, falling 6.7% year on year. It was the first decline since Baidu went public in 2005. Advertising accounts for more than 90 percent of Baidu's total revenue.

At the same time, Baidu's two big bets of 2015, Baidu Nuomi and iQIYI, had both failed to turn a profit.

These factors sent Baidu's market value tumbling through 2016, hovering between USD 50 and 60 billion. Alibaba and Tencent, the two companies long grouped with Baidu as China's internet "Big Three," were each worth around USD 240 billion.

Baidu had long poached executives from foreign firms, but its past hires were usually those companies' China-region heads: people who executed, with no say over business strategy.

Lu Qi was different. From the time he joined Microsoft in 2008 until he left in 2016, he ran an entire division and reported directly to the CEO.

Microsoft has never had more than ten divisions. Each is effectively a small company, with its own engineering, marketing, and finance, and each files its own revenue reports. Their heads are like mini-CEOs, running businesses bigger than most independent companies.

After Microsoft's 2013 reorganization, Lu Qi led the Applications and Services Group, which covered Office, Bing, Skype, and more. Office was Microsoft's most profitable business. In the quarter before Lu Qi left, his division posted a profit of USD 3.1 billion.

Bringing in someone who knew engineering, search, and the management of large organizations to run the company sounded like good news for Baidu.

But open the comment threads under any portal's coverage of the announcement, and all anyone talked about was Baidu's fake medical ads and pay-to-rank search results.

Throughout 2016, Baidu's scandals revolved around healthcare, from the sale of the hemophilia forum early in the year to the "Wei Zexi incident" in May. After that, regulators moved on Baidu's advertising business and its revenue collapsed.

To this day, Baidu is still China's largest search company, and very likely its most technically capable. From its early days challenging the portals' search products, through its long war with Google China, it built an image as a fast-moving, technically inventive upstart. As late as 2006, engineers at Google China running competitive tests still concluded that Baidu indexed new content faster than Google.

Yet barely six years after Google left China, the company's image had been transformed.

Healthcare was the spark. But look back over Baidu's seventeen years and it has always worked the same way: hit a problem, fix the immediate problem, move on. No long-term plan.

In Baidu's early years, when it was racing to catch Google and pouring everything into search, this caused little harm. But once Google left China in 2010 and Baidu took the whole market, becoming a channel that could shape how information moves across the entire Chinese-language internet, the problem surfaced.

To fix a revenue problem, Baidu introduced paid search rankings in its early years

Robin Li has said Baidu's mission is "to give people the most equal and convenient access to information, and help them find what they seek." The line still hangs on Baidu's corporate profile, and it is not far from what Google said at its own launch: "to organize the world's information and make it universally accessible and useful."

For any search engine, the most important thing it can do for users is return the results they are most likely to want.

But when Baidu went looking for a business model in 2001, it let advertising distort those results.

Once pay-for-placement was switched on, the ranking of Baidu's results bent to advertiser money: put simply, the more you paid, the higher you ranked. And back then Baidu drew no line between the results it judged most useful to the user and the results that had been bought to the top.

Only in 2009, when Baidu launched its "Phoenix Nest" advertising system, did its search pages begin, like Google's, to separate ads from organic results.

But the underlying habit of bending search results to serve new business lines never really changed.

After Baidu Zhidao (Q&A), Baike (encyclopedia), and Tieba (topic forums) launched, a search on the relevant keywords would usually surface these Baidu products in the top three slots.

To pull in traffic in its early years, Baidu launched MP3 search

Baidu launched MP3 search in 2003, encouraging users to download pirated music. From the start, Baidu knew this was a grey area; board member Fu Sheng has said they discussed it at the time, but did it anyway.

That business, which fed pirated downloads, helped Baidu grow its traffic and advertising revenue faster. At its 2005 IPO, Baidu said MP3 search accounted for 20% of its search traffic, though some investors put the figure closer to 50% in 2004.

Foreign companies have always had a hard time winning copyright cases in China. Universal, Sony, and Warner sued Baidu over MP3 infringement, and the labels lost.

In 2007, the United States' Special 301 Report called Baidu's MP3 service China's largest pirated-music download site. In June 2008, Universal, EMI, Sony Music, Warner, and other labels issued a joint statement boycotting Baidu and urging their peers not to work with it.

The music service was eventually shut down, only for almost the same story to play out again with Baidu Wenku.

Baidu Wenku collected books and other written works and made them free to users without the authors' consent. In short: pirated content traded for traffic. Fifty writers, including Jia Pingwa and Han Han, went on to publish a "Declaration by Chinese Writers Against Baidu," flatly branding the infringing company a "thief".

Robin Li's reply at the 2011 China IT Leaders Summit in Shenzhen: "On this matter, I've already told my staff to tighten management. My position is clear: if we can manage it well, we'll manage it; if we can't, we'll just shut it down."

After the medical-ad scandals broke, Baidu still solved problems by plugging whatever hole appeared

"Mr. Li searched Baidu for a 'PLA General Logistics Department hospital' to get treatment, spent more than 10,000 yuan, and still failed to cure a minor ailment that 100 yuan could have fixed."

That line comes from a 2008 news report, yet it would sit just as comfortably in a 2016 story.

Medical advertising was Baidu's single largest source of revenue. In Baidu's 2009 annual report, pharmaceutical and healthcare ads were its biggest advertising category, at 10–15% of ad revenue. Next came industrial equipment, education, franchising, and electronics; together these five categories made up roughly 50% of full-year revenue.

In later years the share grew more striking still. According to figures disclosed by Liang Jianyong, the Putian municipal Party secretary, Baidu took in 26 billion yuan in advertising revenue in 2013, and private hospitals from Putian alone spent 12 billion yuan of that on Baidu, close to 50%.

The "Putian network" refers to hospitals controlled by capital from Putian, Fujian, more than 8,000 of them nationwide. They run the gamut: fully legitimate high-end private hospitals, public hospitals bought outright, small private clinics, and contracted departments inside public hospitals. The pattern of patients following online leads and spending a fortune on minor ailments largely points back to this network.

In 2008, fabricated medical ads were exposed by several media outlets. Around the same time, CCTV's News 30 Minutes reported that Baidu was manually tampering with search results, maliciously burying some, and running extortion-style marketing.

Baidu handled it the way an engineer handles a bug: find the target and patch it. Here the target was CCTV. Robin Li appeared on screen eight times during the 2009 CCTV Spring Festival Gala. On the first-quarter 2009 earnings call that April, Baidu CFO Li Xinzhe confirmed that most of an extra 40 million yuan in marketing spend that quarter had gone to CCTV.

The same logic returned in early 2016, with the Tieba forum-selling affair.

Early that year, Han Qiliang, who had spent twelve years moderating the "hemophilia forum" on Tieba, curating posts and teaching the community the basics of the disease, suddenly found that his moderator rights and those of his deputy had been revoked. Three days later, the forum's curated highlights began disappearing one by one. The new moderator was tied to a private clinic in Shaanxi that treated hemophilia.

Baidu later backed down, agreeing to remove the commercially installed moderator, and announced that "all disease forums will immediately halt commercial arrangements and open only to authoritative public-interest organizations."

But in an internal email, Robin Li said the growing fallout was the fault of several parties, but "above all the PR response was too slow, which turned this into a major crisis. Public opinion ran completely away from us and our own voice was drowned out."

For a company that all but monopolized PC web search, public opinion had always been easy to control.

During the 2008 Sanlu melamine-tainted milk-powder scandal, Beijing Evening News ran a story about an internet user who had obtained a leaked document from Sanlu's PR firm showing that Baidu had been paid to delete negative information about Sanlu's milk powder. Baidu's response at the time: "We did get a phone call from a PR agency acting for Sanlu with a request along those lines, but we turned it down on the spot." Yet when 21st Century Business Herald later searched for the same article criticizing Sanlu, Baidu returned just 11 results, while Google China returned 11,400.

By 2016, that approach no longer worked. Most information was already spreading through WeChat Moments and official accounts. China's 700 million mobile internet users were getting their information through WeChat, not through the PC traffic Baidu controlled.

By the Wei Zexi incident, patchwork fixes could no longer solve Baidu's problem

In May 2016 came a far graver episode, the "Wei Zexi incident." A 21-year-old with synovial sarcoma trusted medical ads on Baidu, delayed his best window for treatment, and died.

The tragedy was not Baidu's fault alone. The ads had led Wei Zexi to the No. 2 Hospital of the Armed Police Force, a facility that carried both a handsome military pedigree and tertiary-hospital ranking. But the hospital had outsourced its oncology department to the "Putian network," and neither the regulators nor Baidu had done anything about it.

Something similar had happened to Google once. After an investigation and a fine, it clamped down hard on medical advertising. Search a symptom today and you will almost never see an ad.

When the Wei Zexi incident erupted, Baidu, under pressure, cut advertising on a large number of keywords tied to serious illness.

Eight years of patchwork fixes, and Baidu could no longer wave the problem away. A government investigation team moved into Baidu in early May; a week later, the authorities issued remediation orders that restricted the pay-to-rank mechanism and how ads could be displayed, and required medical ads to be pulled.

In July, the government issued the Interim Measures for the Administration of Internet Advertising, setting rules on ad types, content review, ad labeling, and tax rates.

In the end, Baidu's advertising revenue took a heavy hit, falling 1.2 billion yuan year on year in the third quarter of 2016.

Once advertising stopped growing, the cost of Baidu's shallow strategic forays came due

After search succeeded, Baidu dipped into almost every market going. It is hard to say it ever truly committed to any of them.

Take e-commerce, which Baidu entered and abandoned several times over. It built Baidu Youa in 2008, a C2C marketplace launched five years behind Taobao, and closed it in 2011.

Then in 2015 came Baidu Mall, a site that resembled JD.com in both model and design, arriving eleven years after JD.

There was also finance. In 2008, Baidu launched a payment tool called Baifubao, a name close to Alipay's and one of China's earliest forays into online payments; it was later renamed Baidu Wallet. By the end of the first quarter of 2016, Alipay and Tenpay (owned by Tencent) together held more than 90 percent of China's third-party mobile-payment market by transaction volume. Baidu Wallet had 0.4%. In 2013, Baidu also launched "Baidu Baifa," a wealth-management product promising an 8% annualized yield. The high yield lasted only a month, after which Baidu kept rolling out products at various rates, "Baidu Earn 180 Days," "Baidu Earn 365 Days," and more. Baidu's wealth-management arm had a mobile app, but in Questmobile's ranking of the most active wealth-management apps in spring 2016, Baidu's offering was nowhere in the top fifteen.

Google, too, launched a sprawl of products and racked up plenty of failures. But in a few decisive areas, it held on. Worried that Microsoft would corner the phone operating system and shut out its services, Google pushed Android to market and built it into the world's largest operating system. It also built an entire account system around Gmail, knitting together its full suite of internet services. These moves let Google keep growing today without leaning on medical advertising.

In 2015, Robin Li said, "Baidu still has more than 50 billion yuan in cash on hand. Let's take 20 billion of it and do Nuomi right." That money went into Nuomi and Baidu Waimai, to capture transactions in offline local services such as restaurant reservations and movie tickets. Facing skepticism from Wall Street, Robin Li even warned at one point that if American investors did not understand his strategy, he would consider taking Baidu private and bringing it back to China.

But after a year of widening losses, by the 2016 Baidu World conference, Robin Li had stopped mentioning O2O altogether.

The new story was artificial intelligence, a theme Robin Li had returned to again and again at recent Baidu World conferences and at the briefing welcoming Lu Qi.

Search is inseparable from AI, and at the basic level of image and video recognition Baidu has built up real capability. But how much Robin Li himself understands about artificial intelligence is another matter entirely.

At the 2015 Boao Forum for Asia, Robin Li had this exchange with Bill Gates and Elon Musk, at a moment when Musk had just personally backed a project to oversee AI research.

Robin Li: Elon, I understand you recently said that advances in artificial intelligence are like summoning a demon, which set off a lot of heated debate. Baidu's chief scientist Andrew Ng has said that worrying about the dark side of AI is like worrying about overpopulation on Mars: it only distracts AI researchers.

Elon Musk: I know a little about Mars, and I don't think that's the right analogy. The risk with digital superintelligence is that it won't stop at the human level. It will immediately reach the superhuman level.

A better analogy is nuclear research. It can become an extremely dangerous weapon. Releasing the energy is easy; storing that energy safely is very hard.

I'm not against advancing AI, but I think we should be especially careful. If that means it takes longer to develop AI, then I think that's what we should do. We shouldn't rush headlong into territory we don't understand.

Robin Li: Bill, I know you hold a view similar to Elon's, but is there any difference?

Bill Gates: Not much difference. (He then traded a pleasantry with Musk.) For anyone in the audience who's interested, I'd recommend a book by Nick Bostrom called Superintelligence.

Humans have a general-purpose learning algorithm that came from evolution. It runs on an extremely slow computer with very limited storage, and we have to use this ridiculous tool called a mouth to transfer data to other "computers." Every new computer that gets created has to start over from scratch; it doesn't even know how to walk.

We've never implemented this algorithm in software. But once you write it into silicon, it will immediately surpass humans and become superhuman.

The moment this algorithm connects to the internet, it can read every book, magazine, and other piece of information. We have already prepared the content base for a superintelligence.

I try not to push this view too much. But when someone says this isn't a problem, I genuinely can't agree. How can they not see what an enormous challenge it is? (Gates shook his head as he finished, then looked at Robin Li, waiting for an answer.)

Robin Li said, "We all understand this will have a profound impact on society and on the future. And many companies, including Microsoft and Baidu, are investing heavily in artificial intelligence. These past few days I've been reading Walter Isaacson's The Innovators…" (The conversation moved on to the next question.)

YouTube

Now, Robin Li says that with operations in Lu Qi's hands, he will spend more of his energy on Baidu's future strategy.

For a long time, handsome advertising growth kept Baidu's problems out of sight. Now that healthcare has dragged its advertising business into decline, the familiar company-wide reflex, fix a problem by staring at that one problem, can no longer reverse the trend. And the time that might have built durable strategic positions was used up years ago, one half-hearted foray at a time.

Lu Qi has now been brought in to run the company. But changing the working culture of tens of thousands of employees, and fighting for advertising revenue against a WeChat that holds half of China's phone users for 90 minutes a day, may already be too late.

About the Top 15 Annual Report

2016 was a turbulent year around the world. Among businesses, the market leaders hurt more than the companies chasing them.

The Top 15 companies QDaily (好奇心日报) follows, the ones leading changes in how we live, had a rough year. And they shared a single common trouble: the ceiling.

Whether building new products (like Apple's watch and car ambitions) or chasing new users and channels (like IKEA moving into e-commerce), these industry leaders have all been searching, urgently and perhaps blindly, for the "next" growth engine.

Breaking through that ceiling has proven harder than anyone imagined. The 2016 Top 15 year-end report is not all good news.

Cover image: Baidu