Ep. 27: Poking at the Hornet’s Nest: Fake Reviews in China Tech

Transcript

We are TechBuzz China by Pandaily, powered by the Sinica Podcast Network.
We are a new weekly podcast focused on giving you a peek into what’s buzzing within the tech community in China. We uncover and contextualize unique insights, perspectives and takeaways on headline tech news that don’t always make it into English language coverage. TechBuzz China is a part of Pandaily.com, a new English language site that tells you “everything about China’s innovation.”

(Y: Ying-Ying Lu; R: Rui Ma; E: Eva Woo)

[00:00] Y: Mafengwo, spelled m-a-f-e-n-g-w-o, which means “hornet’s nest,” sometimes also translated as wasp’s nest, is the tripadvisor of China. Well, not quite, but kinda. You can search for user-generated itineraries to almost any imaginable destination on there, and then book the suggested accommodation right on the website or in app.
R: Why would you name your company after such a creature, you ask? Well, apparently at least inside a real hornet hive, it’s a harmonious Communist society and all the insects share and help each other. And so that’s what the founders of Mafengwo had in mind when they created the company, a place where everyone would upload their travel guides unselfishly and share their best travel tips, as world travelers are apt to do.

[00:53] Y: And for a while, it seemed to work. After being founded in 2006, Mafengwo has had a slow but steady climb and now boasts 100mm users, 920K international hotels, and 21mm “real reviews” 真实点评 covering 200 countries. A very generous ecosystem of sharers!
R: But then it all changed. Last week, a blogger named Dingziquan who operates the Wechat account Xiaoshengbibi, together with a three-person data analytics firm called Shenzhen Hurui Data published an article with the following title – “Mafengwo, the startup valued at $2.5Bn USD, is actually a ghost town constructed on zombies accounts and fake reviews?”

[01:40] Y: In it, Ding & Hurui alleged that a whopping 18mm of the 21mm so-called “real reviews” were fake. That’s more than 85%! Immediately, this became the top headline across Chinese tech news for the week, and it is still ongoing, with Mafengwo having already filed a lawsuit against Hurui and Ding for defamation.
R: In this episode, we are joined by guest co-host Eva Woo, who will attempt to unravel this hornet’s nest — haha! total pun intended — with us, and take a look inside the world of fake user generated content in China. How does it happen? What are the consequences? Is it the industry norm? Let’s find out!

[04:35] R: So before we start, can you introduce yourself, Eva, to our Techbuzzers tuning in?
E: Hi ladies! Good to be here. I’m Eva Woo, that’s spelled W-o-o, and I’m a former journalist with SCMP, Hong Kong Newspaper, Bloomberg and Caijing previously covering Asia economic and business news and now living in San Francisco Bay Area. I’ve known Rui quite a while now, since we were both living in Asia actually, and found that we had a shared passion for telling the biggest Asian innovation stories of our time. I’m really excited to be on Techbuzz to tell the story of Mafengwo and fake reviews.
Y: Welcome, Eva! I know this story didn’t get too much attention in the English media, because let’s face it, at $2.5Bn, Mafengwo is a really minor Chinese unicorn, but it stirred up a storm of controversy in China, including, as we’ve said earlier, a lawsuit, and I’m so glad you’re here to talk about it with us.

[05:35] R: I agree, as an entirely Chinese-facing website with no international users, it makes sense that this wasn’t a big deal in the West, but it really belies a far bigger problem of content and data integrity in Chinese tech. So Eva, tell us about the problem.
E: Ok guys, imagine you are on Tripadvisor and you stumble upon a review that starts like this: “Translated by the Concise Chinese-English Dictionary: [colon] For our 5th anniversary, my foodie husband and I decided to go on a wine tour of Tuscany …” how would you feel? You probably wouldn’t trust the authenticity of that review, nor the quality of the winery. Even worse, you might seriously question how much you should rely on that website to guide you toward finding you the right destinations for your travel needs.

[06:28] Y: Uh, a review that starts with “Translated by a dictionary”? Of course not! It just screams “this is fake”!
R: Well, that’s basically what China’s Tripadvisor AKA “Mafengwo,” is being accused of. That review you just heard is representative of many on Mafengwo’s website that was allegedly generated by paid posters, many of which are bots.
E: Basically it all started when someone at data startup Hurui got food poisoning from takeout. They gave the restaurant a bad review, but were then accused by the restaurant owner of being a malicious paid poster. They were so pissed that they decided to develop a product to detect fake reviews, and quite randomly picked Mafengwo as their first target, because you know, according to them, their ads were everywhere, so they thought that Mafengwo must be very popular, and must have a lot of data with which to train their algorithm. Super ironic, right?

[07:25] R: It’s true. Mafengwo had spent over $20mm in advertising for the recent World Cup, so their ads were indeed everywhere. Little did they expect though that those ads would bring upon them this detailed forensic investigation by Hurui.
Y: Hurui’s investigation showed that of the 21 million user reviews on Mafengwo’s website, 85%, or 18 million of them, were fake, and not the quote-unquote “original” reviews the company claimed to have. According to them, 7454 fake user accounts were involved in plagiarizing and faking content, some copying and transferring content directly from rival sites including Ctrip, elong, Meituan-Dianping, and even the English ones we said earlier like Agoda and Yelp using online translators.

[08:20] E: In some cases, the bots were so poorly designed that they weren’t even smart enough to take out the lines “Translated by an online dictionary,” which began the review we heard earlier. In fact, many of Mafengwo’s reviews have very visible traces of the sites where they were ripped off from, including direct references or URL fragments to Yelp, for example, because Chinese tourists are increasingly going global, as you know. I mean, it’s honestly amazing no one wrote about this before, the transgressions are so obvious.
R: Yeah, the fake users, for example, left a comment trail with conflicting identity information in terms of gender, time, and location. For example, In one review they would have a husband, the next, a girlfriend. Well, actually I’m not sure why that’s not possible, hah, especially in San Francisco. But OK this next example is really bad. Sometimes, the bot even grabbed the text of the advertisement that was next to the review at the time of screen capture for some truly laughable results — like in one, you can see text for a private investigator in one of the reviews.

[09:33] Y: And you know what else was suspicious? Mafengwo users were overwhelmingly more active in certain periods during certain years of the company’s decade year history than others, plus, large numbers of reviews were being posted during work hours as opposed to on weekends, when most other review sites get their user reviews, as you would expect! Hurui postulated that some of the fake accounts were actually owned by Mafengwo staff. This are the same staff, by the way, who seemed to also “win” most of the special prizes given out on the site.
E: These shocking accusations were well documented. The original post is full of screen captures that highlight the problematic pages. Maybe because it was chock full of evidence, it quickly went viral on Chinese internet. As expected, Mafengwo PR quickly dismissed it as a malicious “organized attack” full of exaggerations. By the company’s own count, less than 3% of total content was problematic, since reviews only account for a small fraction of total content on the site. There are blogs, itineraries, forums, videos, etc. They also blamed online marketers for the forged reviews.

[10:48] R: Mafengwo had to act quickly and aggressively, because this expose literally came at the worst time possible. I mean they’re in the middle of fundraising. It’s true that the company is already backed by some of the best investors in the business, including Temasek, Hillhouse, General Atlantic and Capital Today, but this August, it was leaked that Mafengwo was in the middle of raising $300mm at a valuation up to $2.5Bn. Again, it’s not a Bytedance by any means, but not shabby either.
Y: And in Mafengwo’s business, it means that its reviews are super important, which is why this article was not just devastating in terms of timing, but also generally the content.
E: That’s right. Mafengwo, as you probably have understood by now, is an online social travel platform well known for its user generated content and social engagement features. It also offers travel booking services such as flight and train tickets and hotel reservations. Advertising and commissions from transactions are their two main revenue streams. At the end of last year, cofounder Lv Gang said about half of the company’s revenue is from advertisers, and the other half from being a marketplace, selling flight tickets, hotel and travel products.

[12:08] R: And just how much revenue was there? According to Jingtravel, in 2017, Mafengwo had $1.5 billion in sales, 63k transactions, and over 100 million monthly active users.
Y: But it’s still a second tier player in the gigantic Chinese online travel market. According to Analysys Yiguan, in Q1 2018, Ctrip and Ctrip-backed Qunar account for 53.6% of the market while Alibaba’s Fliggy or 飞猪 was about 15%. Mafengwo is behind all of these guys, then again the market is huge.
E: That’s because Mafengwo plays in a smaller niche. While Ctrip and Qunar are your traditional, mainstream travel booking sites focused mostly on search, Mafengwo and Qiongyou use user-generated content to drive traffic.

[13:05] R: Which explains why the often-cited 21 million reviews are critical for Mafengwo to demonstrate the active user engagement on its website. So those reviews are also how Mafengwo can charge businesses for advertising, especially those with unique assets, such as boutique hotels, who might not do as well on Ctrip’s one-stop-shop platform.
Y: Mafengwo understands that the user’s first step of planning any trip is not the actual booking process, but the decision of where to go. Which is why it’s content-driven and started with traveler’s blogs.
E: Honestly, in the past I’ve stumbled upon many Mafengwo blogs while doing travel planning online, and some of them are quite good. Most of them seem to be authentic, they have a voice, they’re informative with lots of tips, and tons of pictures and videos attached. Many of my Chinese friends living in China agree, they do regularly use it to come up with itinerary and destination ideas.

[14:07] R: However, such content is incredibly hard to scale and monetize. Which is why Mafengwo eventually transitioned into more of a review business, hence our comparison of it to Tripadvisor.
Y: And, it turns out reviews are also hard to grow within a short amount of time, and under the pressure of scaling quickly, it appears Mafengwo management found a more “creative” way to make its KPIs go up and to the right.
R: And now we are back to where we started, which is this controversy of fake data, namely, fake reviews. Now that you understand why the reviews are so important to Mafengwo’s business, it makes sense why the company is fighting back so hard. As we mentioned before, it’s already filed suit against Hurui and Ding for defamation in Beijing, accusing them of a malicious and organized attack.

[14:58] E: But obviously the defendants are not giving in. Ding Ziquan put out an official statement that their investigation was purely organic and not funded by Mafengo’s enemies as they insinuate. Ding also pointed out that Mafengwo has been hard at work erasing the evidence, although you can find the fake reviews easily via web archive. Hurui pointed out that even with Mafengwo’s core blogs, at least 7% are suspected to be infomercials by paid posters. As Ding put it, plagiarizing and paid posting are so rampant on Chinese internet that it has almost become “a consensus, or an unspoken rule.” They just made the greater public more aware of it.
Y: Yeah, plagiarizing and paid posts are both pretty self-explanatory but let’s talk a bit about the latter anyway. In China, paid posters are called 水军，which technically means “navy.” Why navy? Well, It comes from the fact that many navies in the old days were basically mercenaries for hire, and that’s what these guys are — posters for hire — you pay them some nominal amount of money, like $1 or less, and there are thousands or tens of thousands of them ready to do whatever you need.

[16:21] R: They’ll do anything, like write fake reviews or comments, repost stuff, report and flag posts etc., etc. While in the early days these were mostly done real people, nowadays more and more they are the work of bot farms. You guys think Twitter and Facebook are the only ones with a fake account problem? Think again. The Chinese government, by the way, has been contemplating laws to make such activities illegal, but I think it’s probably going to take a while to enforce.
E: They’re also called the navy because this is how you flood a site with information … or disinformation. And it’s cheap to do so. Just out of curiosity, the other day I did a random search on Taobao with the keywords “Mafengwo+dianping (review)”, it yielded at least 5 travel content marketers. For $150 a month or less, vendors will write infomercials guaranteed to prominently feature travel blog on Mafengwo. There are also Mafengwo accounts for sale, as well as Python web scraping services and review writing services (many of which are AI powered, btw) available for less than $1 per review. Infomercial blogs, not just for Mafengwo but other travel sites, have higher price tags ranging from $50-$80 per piece.

[17:36] Y: In fact, the plagiarism and fake reviews are so rampant that it’s one of the main reasons why according to one report, why none of the other Chinese travel sites have come out and sued or even complained about Mafengwo, because all the other travel sites do it too! In fact, they all lift content from each other so freely that this is almost a “victimless crime.”
E: And it’s not just the travel sector, by any means. Bytedance’s Toutiao has gotten in trouble for copyright issues, being accused and sued by news media and internet aggregators like Tencent, Sohu, and Weibo. But Toutiao has been a victim too. Most recently in May, Bytedance’s CEO accused Tencent’ short video app Weishi of “borrowing” video content from Tik Tok. The video plagiarizing was so popular that there are online tutorials teaching people how to scrape video content from one site to another without being noticed.

[18:33] R: So what you are saying Eva, is that the actual scale of the scandal might be “controversial,” — I mean , 85% fake reviews is a lot, but the fact is, many internet industry insiders in China don’t think it’s such a big deal. And that’s why Ding Ziquan is so upset right? Because this “controversy,” to many, is not controversial at all. It’s become an accepted practice to fake content.
E: My opinion is that when investors only focus on numbers over the quality of the business, and certain KPIs equate to $ signs in fundraising, plus there’s no way to scale those numbers as quickly as the hungry founders and investors want. Combine that with lack of proper due diligence, and a low price for misconduct … then is it any wonder that entrepreneurs decide to be “creative”?

[19:25] Y: Indeed, the upside is much higher than the downside! At this point, I’m afraid that hiring paid posters and copy pasting plagiarized content are probably original sins almost every entrepreneur in China has committed.
R: And how much do investors know? Are they oblivious? Or are they part of the game helping internet companies like Mafengwo dress up their numbers before going for fundraising and IPOing in the US? That’s not my conspiracy theory, that’s what Chinese netizens are wondering. I mean, Mafengwo has actually put out an official statement that they don’t expect the data controversy to affect their plans of going IPO on the NASDAQ at all. They still expect that to happen within the next 2 or 3 years. I’m not joking.

[20:15] E: Well, maybe the investors don’t know, because at least for Mafengwo, there was a dramatic growth in user activities during the times when the company was fundraising. So it could be argued that the data was faked in preparation for the fund. Either way though, it seems like maybe investors should hire firms like Hurui to dig deeper into prospective investments because I’m sure there are plenty more skeletons to be found.
R: And it may be good for business too, because at least anecdotally, it seems that millennials in China actually care about the authenticity of the product they are using. And they want to connect with other passionate users — not bots. While it is true that there are many in China who take a really utilitarian view — like, those people who believe that, oh whatever, even if these reviews are plagiarized they’re somewhat useful to me — those people they tend to be older, over 40. Whereas the younger folk do not think it is cool at all. So maybe changing user behavior can drive some of these entrepreneurs to think twice before engaging in shady practices.

[21:28] Y: Eva, before we end, what are your final thoughts on this story?
E: I really liked working on this episode with you guys because it shines some light on how these Chinese internet companies are managing to grow so rapidly to make their investors happy … that is, with the unfortunate attitude that the ends justify the means. It also makes me wonder how much we could really trust some of the phenomenal user data reported by these Chinese internet companies …
R: I know, it sounds like we need more people like Ding Ziquan and the folks over at Hurui to keep these companies in check. Ding, by the way, was the same guy who exposed the Chinese Redcore browser for being just a skin of Google Chrome a few months back. I don’t know if you guys remember that scandal. Hah. This guy is fearless! Hopefully knowing that more and more people are watching and paying attention will keep more and more entrepreneurs honest.

[22:23] Y: Well said, Rui! And investors too, they should be more demanding and rigorous in their due diligence. Anyway, there definitely needs to be repercussions, because it’s the consumer who’s hurt and cheated out of their hard-earned cash at the end of the day.
R: Alright, One last thing before we end, everyone. How can Techbuzzers find you, Eva?
E: Probably easiest to reach me on Twitter at my name, which is spelled @evawoo, and my linkedin name is the same Eva Woo. Thanks everyone!

TechBuzz China by Pandaily is powered by the Sinica Podcast Network. Pandaily.com is a new English language site that tells you “everything about China’s innovation.” You can find us on twitter at @techbuzzchina and @thepandaily, or reach out to Rui and Ying-Ying at @ruima and @ginyginy. Our producers are Bonnie Zhang and Kaiser Kuo. Our intern is Wang Menglu.