Episode 12: Yinglian Xie & Fang Yu of DataVisor: Fighting Frauds with Machine Learning

GGV Capital’s Hans Tung and Zara Zhang interview Yinglian Xie and Fang Yu, the co-founders of DataVisor, a fast-growing startup in Silicon Valley that provides big data security analytics for consumer-facing websites and apps. Its customers include some of the largest companies in the world, such as Alibaba, Dianping, Pinterest, Yelp, Bytedance (a.k.a. Toutiao), among others.

Both Yinglian and Fang have decades of experience in Internet security, specifically on fighting large-scale attacks to online services, such as fraudulent online payments, spamming, user hijacking, search-result poisoning, etc. They were both senior researchers at Microsoft for many years before starting DataVisor in 2013, and have filed over 20 patents.

Yinglian received her Ph.D. in Computer Science from CMU and a Bachelor’s degree from Peking University. Fang holds a Ph.D. in computer science from Berkeley and a Bachelor’s degree from Fudan University.


HANS TUNG: Hi there. Welcome to the 996 Podcast, brought to you by GGV Capital and co-produced by the Sinica Podcast. On this show, we interview movers and shakers of China’s tech industry, as well as tech leaders who have a U.S.-China cross-border perspective. My name’s Hans Tung. I am the managing partner at GGV Capital, and have been working at startups and investing in them in both the U.S. and China for the past 20 years.

ZARA ZHANG: My name is Zara Zhang. I’m the investment analyst at GGV Capital and a former journalist. Why is this show called 996? 9-9-6 is the work schedule that many Chinese founders have organically adopted. That is, 9 a.m. to 9 p.m., six days a week.

HANS TUNG: To us, 996 captures the intensity, drive and speed of Chinese Internet companies, many of which are moving faster than even their American counterparts.

ZARA ZHANG: On the show today we have Yinglian Xie and Fang Yu or Xie Yinglian and Yu Fang in Chinese. They are the cofounders of DataVisor, a startup here in Silicon Valley that provides big data security analytics for consumer facing websites and apps. Its customers include some of the largest companies in the world such as Alibaba, Dianping, Pinterest, Yelp, ByteDance or Toutiao among others. Both Yinglian and Fang have decades of experience in internet security, specifically on fighting large scale attacks to online services, such as fraudulent online payments, spamming, user hijacking, search result poisoning, etc.

HANS TUNG: They were both senior researchers at Microsoft for many years before starting DataVisor in 2013, and have filed over 20 patents to date. I first met both of them in 2014 and they have come a long way since then. Yinglian received her Ph.D. in computer science from CMU, and a bachelor’s degree from Peking University. Fang holds a Ph.D. in computer science from Berkeley right here, and a bachelor’s degree from Fudan University in Shanghai. Welcome to the show, Yinglian and Fang.

FANG YU: Thank you Hans and Zara.

YINGLIAN XIE: Great pleasure to join you. Thank you.

ZARA ZHANG: First of all could you give our audience an overview of what DataVisor does in layman’s terms, and describe how you’re helping some of your most important clients today.

YINGLIAN XIE: Sure, I would be happy to. DataVisor is a company that leverages the latest technologies such as AI, machine learning and big data to protect consumer facing companies from a variety of fraudulent, abusive, and money laundering activities. Some examples of those attacks include what we are familiar with, the fake accounts, which are a very hot topic. They could introduce spamming like sort of content and sometimes fake content. There could be fraudulent listings and scams. These are the familiar terms.

HANS TUNG: They’re familiar to those of us who operate in China or spend time in China, but for the Western audience, why are there so many fraudulent listings and accounts and so forth? What’s the purpose?

YINGLIAN XIE: Everything migrates online and we have this era of going to the Internet and mobile app era, and the opportunity of directly touching a large number of users from our perspective has significantly enhanced and expanded. Those channels are at the same time being quickly noticed by the attackers to facilitate all these illicit activities we see. Those included, for example, some that we talked about like spam contents and fake listings, and in this case these are professional attackers, not just one-off things. These are actually coordinated organized crime rings and they have a strong financial incentive.

HANS TUNG: How do they make money with fake listings?

FANG YU: There are plenty of extra ways they can monetize through fake listings. For example, they can actually set up some fake listing to attract normal users, so that they can make money through scams afterwards, or more often they also set up a business and trick some small business like, “I can promote your business if you pay me this much, and I can help you find a lot of fans, I can help you with writing reviews”. And then they can also do other things, for example, they can try to hijack some existing people’s accounts and then they can put up the fake listing there, so that when people click those links they can actually download malware so they can monetize further through other channels.

ZARA ZHANG: So a lot of your customers are consumer internet companies like Alibaba or Dianping, the ones I mentioned. Do you see more of those attacks in China or in the U.S., and how are your clients’ needs different in these two markets?

YINGLIAN XIE: In terms of attacks we see is a common issue actually across both China and the U.S., and that’s where, for example, DataVisor is seeing customers in both regions because they all face common challenges. We talked about these fake accounts as one example of these attacks, and what DataVisor protects is a broader category of attacks, for example, including fraudulent transactions which is maybe more familiar as credit card leakage, so when the data gets leaked and someone obtains this information, then they just this credit card to issue a payment transaction, without the owner actually knowing about the existence of the transaction, which results typically in what we call chargeback attacks on merchants and on banks.

And there are also other types of attacks such as fraudulent installs. Fraudulent installs are more in the marketing world, and there are many apps that are looking to expand their user base by, for example, acquiring users from different channels such as Facebook or Google, and there are many other channels where they could acquire users as well. And in different kinds of channels sometimes there are also sub-channels and sub-networks, and there could be places where, for example, the purchases are actually fraudulent, meaning they never existed. These are not really user installs, and this is another example of fraudulent activity that we see very often. There are other types of use cases.

We are familiar today with identity theft. Our sensitive information such as social credit card information, our social security numbers, our credit card information, our address, these are sometimes being leaked. That means the identity can potentially be abused by someone else that we don’t know about. They use this information to apply for new credit cards, and also, for example, to open new bank accounts without the real user actually knowing about their existence. When they make these new credit cards, new accounts, then they can make transactions and other types of illicit activity we talked about.

So these are different forms of attacks that we saw that actually happen in China in consumer facing applications and online services as well as those in the U.S. commonly. Obviously fraud patterns might differ slightly in these regions, but it is a common challenge. This is something that happens with all these online services quickly increasing, and we are actually moving from offline era to the online era, and it just makes these attacks far more easier than in the past, and these are the problems that DataVisor is trying to address.

ZARA ZHANG: How did you two first meet each other?

FANG YU: We met each other at Microsoft Research when we first joined Microsoft. We two actually joined the same month the Microsoft Research Silicon Valley lab, and we’ve been collaborating with each other ever since then.

ZARA ZHANG: Which year was that.

FANG YU: That was in 2006.

ZARA ZHANG: So it’s been 12 years.

FANG YU: Yeah.

ZARA ZHANG: And you were at Microsoft for almost 10 years?

FANG YU: Less than that.

YINGLIAN XIE: We worked together at Microsoft for seven years.

ZARA ZHANG: Did you guys work on projects together while you were there?

YINGLIAN XIE: Yes, we closely collaborated with each other since the very beginning, and we worked on many of the projects in the broad security area.

ZARA ZHANG: So was it a difficult decision to leave Microsoft and when did you know it was the right time to make the move?

FANG YU: Maybe I can talk about it. I would say from that time, when we made that decision, it was very interesting. I know in the Silicon Valley a lot of people are thinking about startups a lot of the times. So that idea actually didn’t occur to us for a very long time, but when the idea occurred to us it was actually quick and easy for us to decide that’s something we are very interested in doing, and that is something exciting that we wanted to do. I would say from a decision perspective it wasn’t a really hard decision.

HANS TUNG: I remember Harry Shum from Microsoft was the one who connected us. You guys had already left Microsoft and started doing this. How did you guys know Harry and how did you get him to believe in you guys?

YINGLIAN XIE: Both my husband and I came from the CMU ward and Harry is a legendary senior alumni, and a role model coming from CMU.

HANS TUNG: That’s right. He is essentially the defacto CTO of Microsoft and runs AI?

YINGLIAN XIE: He is the CTO of AI, that’s right. He has been the role model for us and all CMU graduates, particularly in this area of AI and machine learning. Harry got to know us actually before our Microsoft time when my husband graduated and left CMU, and he was trying to recruit us to join the Microsoft Asia lab. That was the early days..

HANS TUNG: That’s based in Beijing?

YINGLIAN XIE: That’s right.

HANS TUNG: Started by Kai-Fu Lee?

YINGLIAN XIE: From that moment on he’s been like a mentor to us and a role model. So when the two of us actually left Microsoft we went to talk to Harry about our idea and what we wanted to do, and he gave us a lot of encouragement and help in brainstorming this, and talking about this. We keep in touch from time to time and we’ve got a lot of advice from him.

HANS TUNG: How did you guys decide that this was the problem that you want to solve?

FANG YU: We have been working in this area for a long time, ever since our Ph.D. My Ph.D. was on network intrusion detection, and Yinglian was actually working on forensic analysis, so we were always in the network security field. And in Microsoft we are a little bit higher up because we don’t do network level stuff a lot in Microsoft. At Microsoft we worked with many product groups although we are in research. We worked with Bing, we worked with Hotmail, Forefront, and then Skype, etc., so we worked with many internal teams. At Microsoft we developed a lot of a systems to help these internal product teams to fight virus type attacks.

The techniques we did there were very focused, so when you have one problem, we have one specific solution for you. When you have another problem we have another brand new solution for you. After seven years we felt that everything at the surface is very different, but everything at the root is actually connected. Instead of chasing them one after the other, when you look at the way they behave you chase them, it’s already late. So we figured out that actually there is a lot of commonality in there, and the people attacking Hotmail tomorrow may be attacking Bing. So they are all connected at the root, and that’s why we felt we want to leave Microsoft and build a generic solution that’s no longer project based, not chasing after things have happened.

We talked to different people and we saw the opportunity, because most people even today, their solution is actually after the fact. They see an attack, they’re building rules. They see another attack, they have labels and start training models. They are always very reactive, and that wasn’t something that we wanted to be, we want to be proactive and look at things automatically to discover new patterns. That’s the idea we had. We wanted to build that and that’s why we left Microsoft to give it a try.

HANS TUNG: Got it. Was it easy? How difficult was it to do your own startup having been in the comfort of Microsoft for such a long time, because for all our listeners, it’s not easy. Even we know there’s an idea and it makes sense, and there’s potentially a part of market for it, just take that first jump, that leap of faith to leave a very comfortable, successful career in a safe company to come out and do this is not an easy decision. What was that like after you made that leap?

YINGLIAN XIE: I think from of the decision perspective it seems like a relatively easy decision in retrospect, but the journey was not easy.

HANS TUNG: Everybody wants to know about your journey.

YINGLIAN XIE: And talking about why this decision seemed easy is I think after being this area for a long time both of us had this passion that we wanted to do something bigger.

HANS TUNG: Because you see all the unicorns being built?

YINGLIAN XIE: We were in the research world and Microsoft Research at that time was kind of focusing on advancing the state of art, and that was we were actually doing, and we were very good at it. But at the same time we saw this gap between research and industry, and the two value propositions are very different. So at the same time it’s very difficult to get these things across particularly in the security area. So we were very excited coming from research to see can we bring this new ideas and latest technology, whatever we come up with, quickly to production. And that’s something brand new and sounds really exciting.

We think also the timing for us worked very well, because the mega timing is everything moving online, that made attacks a lot easier, and everybody was in need of a such solution. On the other hand we thought that machine learning, AI and big data as a technology evolved to the stage where it was more mature to productize it and bring it to the market. So we were very excited about that possibility. Leaving the comfort zone definitely is something a lot of people would question, but I think in this case we two had some advantage of being ladies, because to some extent if we were guys we would maybe feel the family burden a lot more, and you’d think twice about leaving a stable job earning the family money. We were both very thankful to our husbands at that time as they were both working.

HANS TUNG: Are they still working?

YINGLIAN XIE: Yes. They were also supporting the family. So we kind of went to our husbands and brought up the idea. There was this joke where I went to my husband and told him about this and saying, “Well, why don’t you treat me as a housewife except I don’t do housework.” My husband’s reply was, “Well, you were never the housewife type and now you even want to earn money.” I guess that also is another factor to some extent, and we feel there were some advantages in the family support, and being ladies we felt a little bit less pressure.

FANG YU: We were very thankful to our husbands.

HANS TUNG: Yeah. Same thing with you, same conversation?

FANG YU: Yeah, I think my husband was actually very supportive.

ZARA ZHANG: You both had young children when you started the company, right?

FANG YU: That’s right. Very young children.

HANS TUNG: So how did you deal with that?

FANG YU: I have nannies so that actually gave me a lot of help to actually support me doing the work.

HANS TUNG: If you can give me any recommendation, my wife and I are still looking for one?

FANG YU: I’m very thankful to my husband, my parents and in-laws, the nannies especially.

HANS TUNG: So your parents live nearby too?

FANG YU: Yeah. My mom lives nearby.

HANS TUNG: Huge difference.

FANG YU: So I think we got the whole family support.

YINGLIAN XIE: Similar here. My parents gave me tremendous support. They stayed with us and supported us, and that’s why we just said we are very thankful for family support. I think this is the same with anybody who does startup, whether you’re a woman or a man you need family support.

HANS TUNG: Right, because your schedule is very long and you’re always at work, so you need someone to back you up at home. So the decision was easy, and the support was there, so what’s the journey like after you start a startup? How difficult is it to get stuff off the ground, hire people, and then try to convince people to join you, because DataVisor didn’t have any brand at all, and you two hadn’t had any successful startup before? So how easy was it or how difficult was it to get the project up and running and build the first version of product?

YINGLIAN XIE: When we actually looked at this we thought the decision was easy, largely because we didn’t really know what we’re getting ourselves into. Before we made the decision we talked to colleagues and discussing this. They were giving us good advice and pointed out entrepreneurs we should talk to really see are we ready, and is this something we want to do to make sure before jumping to this. So we did that. We actually talked to a few folks who have been doing startups about what does it take to do a startup.

And after that we came back and we were asked what do you feel, and we both had the same answer that we weren’t ready. But that didn’t affect us and we still wanted to do it. There is always somewhere to get started and we have to start somewhere. The first step was to quit our job and start really working on it. There were a lot of things we truly felt we weren’t ready for, and a lot of learning in our way, but we decided to do it, and then that’s where we started. I would say the journey is definitely harder than when we started initially, perhaps because initially we didn’t have any expectation on what it’s going to look like, so pretty much no expectation, and it was actually not easy.

You talked about for example recruiting. My husband’s working in Microsoft and leading a large team as well, and everybody’s complaining about recruiting being very hard in the Silicon Valley, very competitive. But when you were in a startup you truly feel how much harder it is. When you’re in Google or Microsoft recruiting that’s very different than a startup.

HANS TUNG: So how did you guys get the initial hires? How did you get the first two or three hires to join you, to believe in you.

FANG YU: I think the first few people were people that we knew from before, and then we hired those people because we knew they’re great people to actually start that journey with. And then moving on we had many more, we actually started to have referrals, and then going on we started campus recruiting, and we started linking programs, we hired internal recruiters etc. So then we got the ball rolling to grow the company.

YINGLIAN XIE: It’s a lot of hard work and strong discipline. Initially that meant we grew slower, but one thing we did was hold our bar of recruiting high, and I think over time it paid off, because we set up a stronger team. And when you have a base it actually makes the further down recruiting slightly easier.

ZARA ZHANG: How many people do you have?

FANG YU: We have already I think over 80 people today and we’re quickly growing.

ZARA ZHANG: So both of you grew up in China, went to college there, and then came to the U.S. for grad school, so you still have a very strong connection there. A lot of your current customers are based in China, so have you thought about starting this company in China, and why did you choose to start it here?

FANG YU: It was pretty natural to start it here because we worked in Microsoft and our homes are here. And also it’s Silicon Valley, right? It’s a place of innovation. So we didn’t think of starting in China until we have a good customer base. We started the China branch in 2016.

YINGLIAN XIE: Particularly because we’re working in the technology field, Silicon Valley is this field for both startups and high technology. So it didn’t occur to us when we have a startup that we need to find different places, because when we started DataVisor the idea was it’s a technology driven startup.

HANS TUNG: Now do you have engineers in China?

YINGLIAN XIE: Yes we do.

HANS TUNG: Was it easy to who those engineers as a U.S. company and as a sea turtle returnee?

YINGLIAN XIE: I would say initially we thought we have this Silicon Valley angle and it must be somewhat attractive.

HANS TUNG: There are so many successful companies in China so recruiting is hard.

YINGLIAN XIE: Right, there are so many successful companies in China and I think they grow really fast as well. It was also very difficult recruiting there.

HANS TUNG: There are more consumer companies than enterprise companies, and consumer companies grow even faster, so the attractiveness toward enterprise companies is actually less in comparison, so it makes your job even harder.

FANG YU: It’s hard.

HANS TUNG: So how did you overcome that? Did you go through the Microsoft connection to meet more people that know you, and are therefore willing to try and join you? What was your hook to get the first few engineers in China to join?

YINGLIAN XIE: First of all we had one of our early founding members who actually led the China team. He went to China, relocated. That’s actually a big step, someone being there.

HANS TUNG: How long had he or she been living in Silicon Valley before relocating back?

FANG YU: He was in the U.S. actually for quite a few years already.

HANS TUNG: Four or five, but not like nine or ten?

FANG YU: Yeah, probably.

HANS TUNG: Because nine or ten it would be harder?

FANG YU: Yeah, very difficult.

HANS TUNG: Four or five is still possible?


YINGLIAN XIE: Going to China was a big first step. We did locally get help from our investors there. These were actually early anchor points, some of the clients, some of the early recruits, referrals.

HANS TUNG: I even made some introductions for you, even not being an investor myself.


YINGLIAN XIE: And toward the later stage when we wanted to start product development in China we had a few very capable team members from the U.S. who wanted to relocate back to China, and that helped the team tremendously as well.

HANS TUNG: People who were willing to move back, otherwise would have to stay here for too long.

YINGLIAN XIE: They did similar work here for some time, for a few years.

HANS TUNG: But not too long?

YINGLIAN XIE: Not too long. And to some extent we found that that opportunity was attractive to them as well.

ZARA ZHANG: So who was your first customer and what was the process like attracting them and onboarding them?

FANG YU: Momo was actually our first customer. We were actually just closing the first funding round and then we met the Momo’s founders by accident. One of our friends who helped them tour around Google actually introduced us to them. They talked to them and asked, What’s your biggest problem, and they said spam problems. So he said, “Oh I want to recommend our friends actually who have started a startup.” So we talked to them at Starbucks close to Stanford at night, and we talked about the product we are building, the initial prototype that we have.

They actually got pretty excited about all solutions and they trusted the two of us and our small team that we can help them solve the spam issue. Then we actually went back to do PoC with them and those results were very good, and we quickly moved on board. So getting our first client was actually very lucky and we happened to meet him.

HANS TUNG: Is Momo a paying client?


HANS TUNG: That’s interesting because Momo’s market cap back then was a few billion dollars like Yelp is today. Do you think it will be that easy to get into Yelp here?

ZARA ZHANG: To give people more context, Momo is sort of like the Tinder of China, it’s a social networking, mostly anonymous social networking for people to meet new people.

HANS TUNG: Maybe Yelp is not the best example, maybe we can compare it to Snapchat, so can you imagine meeting anybody senior at Snapchat, having a few hours conversation and getting this deal?

FANG YU: We still had to do the PoC, so they just gave us the opportunity to prove ourselves.

YINGLIAN XIE: We were lucky we got contact, but on the other hand I think it is also something reflecting today’s world, particularly China. Momo is an example of a company that grew really fast in three and half years it went to IPO, so they grew phenomenally. And their team is also relatively young as well, a young team growing up quickly. Spam problems of that scale became a key issue that they had to solve correct for them to go to the next step. I think that’s a reflection.

There are many companies in China that face similar challenges that have really fast growth and a relatively younger and inexperienced team handling really large scale issues that is very detrimental to both user experience and business growth, and that’s where our expertise could help. So I would say we were lucky to get Momo as a first client but that also like strengthened our belief of going to China as one of the major markets, because there are many more companies like Momo that could benefit from our service.

HANS TUNG: From our experience of investing in so many consumer internet companies is that the market is so big, and the growth is so fast, that they know they have issues they need to solve. And like you said not as many of them have dealt with large scale issues before, therefore they are very open minded to try anything that they know makes sense. So it was very meritocracy driven when they picked you, in spite of a lot of hurdles, and who you know, have you had other references, are you are compliant to this regulation and that regulation. There’s much fewer hurdles for enterprise companies to penetrate the Chinese market.


HANS TUNG: If they know you and they can trust you and so forth.

YINGLIAN XIE: They’re able to take risks and embrace new technologies, and are very willing to try out new things. And I think that kind of spirit helped them really grow fast. At that time when they were our clients they have not gone IPO yet, but afterwards when the spam problem was sorted out, and that issue they had that was giving them a headache suddenly disappeared, I think that let them to focus on their core business growth.

HANS TUNG: And one of the reasons we invited you to come on the podcast is also that we want that a lot of SaaS companies as security solution companies in the U.S. to know that China is actually a great market, but if you don’t have anybody on your team that’s from China to help you to penetrate that you think that it is very difficult and it takes a long time, but for those of you who know both China and U.S., actually China is an even better market than the U.S. to crack and get things done.

YINGLIAN XIE: I think China definitely is a very promising large market. First of all the market size is huge and there are so many opportunities in a lot of internet services. And traditional companies are moving to online at a really fast pace. Another thing I noticed is really their willingness to adopt new technologies and new solutions to help them grow faster. I think that’s a general sentiment we felt.

HANS TUNG: In the U.S. it allowed us to cut costs. In China it’s to help them to make more money and grow faster. Very different motivation. And people usually move faster when they can help them to make money.

ZARA ZHANG: So today how many percent of your clients are in China versus the U.S.?

FANG YU: When we entered China we had kind of a balance with half of our clients being from China, and today we definitely see the China market grow faster, so the number of customers, growing at a faster speed than the U.S. market.

HANS TUNG: Faster scale, bigger scale. And that’s what we’re seeing with other companies that are selling both into U.S. and China. When they have a comparison Chinese markets, Chinese clients are more open, more willing to try new things and grow at a faster pace. But a lot of U.S. software SaaS companies who don’t have the China market as a benchmark don’t know that.

ZARA ZHANG: You guys have a very bicultural team, a lot of team members have a Chinese background and you have an office in Beijing. Is it difficult to run a bicultural team in both countries at the same time? And do you think that’s a long term competitive advantage?

YINGLIAN XIE: I think Fang probably wants to comment on the bi-lingual culture part because from the technology perspective working with the customer there are always day-to-day issues every day. There is a time difference, there’s culture issues.

FANG YU: I think there’s definitely a culture difference between the Chinese customers and the U.S. customers. As Hans mentioned Chinese customers usually move quickly. They’re doing the trials quickly, prepare things quickly, but at the same time they actually have a lot more requests than the U.S. customers, and they want more customization so they have more requests. So supporting both U.S. and China is actually a bit different for us. We have an U.S. support team and also a China supporting team. And even you have all the development in the U.S. the support you actually need to have local, and having face-to-face time with the customer is actually of highest importance in China.

We do actually work with the China team remotely and with the customers as well. That means that we actually join remote calls. But in the meanwhile I think that technology now is actually very good because we can easily hop on a call with our customers, or with our internal members in China, and we can communicate very easily. So we have all the telecommunications advantages to use. Of course, people still need to go visit them from time to time but I think the remote collaboration has been working reasonably well in DataVisor.

HANS TUNG: So you have a lot of calls at night here, and morning or noon time in China?

FANG YU: It’s morning and at noontime in China, yes.

HANS TUNG: Because by the time you get up in the morning China is closing until midnight.


HANS TUNG: It’s very difficult.

FANG YU: Sometimes we talk with our China team in the morning, but it’s only for a few engineers that are actually very late, but usually some of our company meetings will be in the late afternoon to accommodate the Chinese teams.

HANS TUNG: That’s right, to make that work.

YINGLIAN XIE: And we encourage communication across both teams. We will have the entire China team visit our U.S. team, and we also have members from here going to China to do trials to work with them together..

HANS TUNG: How big is your China team now?

YINGLIAN XIE: The China team is probably about 30 people today, but is quickly growing to more than double that number.

HANS TUNG: How many people do you have here by comparison?

YINGLIAN XIE: We have about 50 people here.

HANS TUNG: So quickly you’ll more in China than here?


FANG YU: Hare we are growing too.

HANS TUNG: So probably 50/50 for a while? And are most people in China in Beijing or elsewhere?

FANG YU: We have both the Beijing office and the Shanghai office right now.

HANS TUNG: What’s the rationale of having two?

FANG YU: We first opened the Beijing office in 2016 and now we have quite some clients in Shanghai, so that’s why we opened our Shanghai office late last year, and that’s why that team is growing.

HANS TUNG: So mostly for support and sales?

FANG YU: Support, sales, and then we are hiring a few engineers there as well.

YINGLIAN XIE: We feel also from a B2B company perspective, and given the China culture we want to be with the customer and listen to them, that helps with some of the development as well. We also have R&D teams located in both places, and the reason is that they are able to directly get to know the business values and listen to the customer voice, to be able to perform the support. And I think it also makes recruiting slightly easier because as we talked about the recruiting in China is also difficult as well, particularly in Beijing. I think having a presence of different locations also enables us to grow the team a little bit faster.

ZARA ZHANG: So how does your approach to fraud detection and prevention differ from the traditional kind? What differentiates your technology?

FANG YU: I think our main differentiation is what we call unsupervised detection, which as I mentioned before, is that most of the solutions currently is more reactive. You see something bad, you write the rules, or you see something bad and you start training machine learning solutions. Those solutions actually work well when the things are stable, for example, face recognition or image recognition, what people look like doesn’t change. But in fraud scenario that changes very quickly. Fraud patterns sometimes evolve overnight, or over a few hours. If you’re always writing rules or training machine learning solutions it’s actually very slow and always reactive.

That’s why we developed something brand new which we called unsupervised machine learning solution, that is able to detect things on the fly, and looking at things in real time detecting new fraud patterns. That’s what differentiates us from all the other solutions. Since we started in the social field we are actually built to scale. So we can process billions of events very quickly in our distributed engines and that’s also the main differentiation versus any other competitors that we’ve actually seen in the space.

ZARA ZHANG: Do you hope to expand into new areas going forward?

YINGLIAN XIE: Yeah. Currently we started with the social sector and that’s where with Momo and Yelp. In social networking we see a lot of demand. We are still going pretty strong in that sector. Then we moved into the gaming sector in the middle because a lot of people actually have hundreds of millions of dollars to buy new users in gaming, and many of those are actually fake installs because there is a limited source of good installs, and there are subchannels, and some of these subchannels could be bad. So the gaming companies or whoever needs to buy new users ends up having wasted their marketing dollars, so we help them to actually evaluate all the buys and help them get refund for those bad installs.

That’s the second market that we are actually doing pretty strong in. And then we are also moving to the financial market and we work with some of the largest clients, money transfers, banks, etc., because there are a lot of issues in financial institutions. For example, in the U.S. there’s a lot of things credit based, but when the credit bureaus information leaks anybody can actually pretend to be me to apply for a credit card online, and you can get the credit card and start to conduct attacks. So there are a lot of coordinated attacks. And then we also see a lot of synergy between the different sectors, because many of the attackers who are playing in the social sector think, “Oh, I have found success here, so I can launch a large scale attack.

Tomorrow I’ll start attacking the financial industry and that will give me a much higher ROIs. So that’s why we’re shifting there as well to help protect the financial institutions from different types of attacks.

HANS TUNG: So over time you see similar players that attack different sectors?

YINGLIAN XIE: Very similar patterns. The attack is always evolving and it is very advanced, and they change very quickly. Traditionally people think financial institutions have different attackers than the social sector because spam is actually for each spam it doesn’t actually get even a penny, but at financial institutions if you get one card you are able to have revenue generating. But now that these gaps are shrinking, and because they can easily launch large scale attacks they will try to attack wherever they can get the revenue.

ZARA ZHANG: I know you’ve seen a lot of interesting fraud cases. Can you tell me an anecdote?

FANG YU: It’s really interesting to see some of the fraud patterns nowadays. One of the cases that we saw, for example, last year was actually a big ATO attack. ATO stands for account takeover. Account takeovers before were actually in a much smaller scale, but this attack actually involved millions of accounts who got compromised. The part that was making this very different from the traditional ATO was that it quickly evolved, it actually changed superfast. In the beginning if you looked at dummy ATO attacks there was just through credential stuff, trying to log in and some similar things. But for this particular ATO attack they actually leveraged Botnets, so they leveraged IoT devices all over the world. Each IP address had only very, very little activity, and each one coming from places distributed around the world.

And then after logging into these accounts their attack patterns very quickly changed. They were very aggressive. They were logging in, and starting to do bad things immediately. For the bad things they do immediately, you can write a rule to detect this. You immediately log in a lot of bad things, and this is a signal. That type of behavior got quickly captured, but then they changed it to very stealthy, and they rate limit themselves very quickly, and they did only one thing at a time, and then another thing, and they had a lot of randomization. This change was made in a few day, and then when they saw that this thing doesn’t go well they quickly switched to another pattern, which is they only log in and don’t do much at all, and they randomize the actual doing things.

So you can see that during the whole week they were doing this attack, first they used very sophisticated methods of actually doing it through IoT devices, that pretended to be iPhones, which is the most popular device there, and then they quickly changed patterns. These attack pattern changes was something that we see as very popular among all the attacks these days. With some clients we saw that the attackers were very patient. Initially they would create many different mass registered accounts and use that to attack. And then when they see that it’s not very promising and that our solutions can detect it, then they would just register the accounts and don’t do anything with them.

They’d just wait for a few months and don’t do anything, and then they start to attack, because by that time the rule based system was recognizing these as old accounts and good accounts, and then they start to attack. And then at some point they figured out that if that accounts becomes very socially connected and their profile looks good, then actually the chance of them having a successful attack was bigger. So working like that they quickly built connections with the real users and among themselves, had good profile pictures, had a bio, and attacked that way. And even though they saw the solution was able to capture them that way then they moved to old ATO attacks, and compromised, older user accounts, and conducted the activity there.

So the trend that we are actually seeing is that a lot of changes around the different behaviors, if your solution is to see that behavior and deploy something, then you are always chasing after what they conducted, and it’s always reactive. That’s why I think we have a lot of success in detecting these very advanced and distributed attacks using our technology.

ZARA ZHANG: Yinglian you mentioned it was actually easier for you because you were a woman starting a business, but have you had times where you felt it was harder because you were a woman, and have you encountered instances where people were treating you differently?

YINGLIAN XIE: I think definitely as a woman in Silicon Valley where there is a male dominant right.

ZARA ZHANG: Right, and you were in a very technical field.

YINGLIAN XIE: Right, you would see different views and different perspectives coming up. I think it’s a normal phenomenon we’re seeing today, and that’s just the reality of these different views and perspectives. Overall I think we too were probably still lucky because along our way our families, investors, and friends were all super supportive, despite the fact that the two of us, two cofounders, were female. Nobody put serious doubts and questioned our choice because of that particular combination factor to question whether it’s going to be successful or not. We got tremendous support. But definitely there were occasions where that happened. Sometimes we did sense a different view.

For example a common theme would be we went to parties hosted somewhere by investment firms who sometimes like to invite people to go, and when there are a lot of attendees, guys, and being female there, and of the people you don’t know about, sometimes it’s difficult to start a conversation because the tendency is that guys like to talk to guys. We sometimes felt that. We’d like to see more woman entrepreneurs being there. In fact there are some, there’s a support community for female entrepreneurs. Before we left Microsoft we talked to people by introduction of friends, to ask about the startup journey. We talked to, for example, some investors, just as a casual conversation, and the idea immediately starting from him was, “Oh, seems to me that two charming ladies want to have a startup.” That’s what happened for example. He was subconsciously asking us is that going to be successful, are they serious about it.

ZARA ZHANG: But both of you also have very strong technical records and careers before you did this, so I guess when people see that they’re like, “Okay, they’re serious.”

YINGLIAN XIE: So I would say such things exist, but I think for the most part the people surrounding us, our family and friends are all very supportive.

ZARA ZHANG: So what advice do you have for other female founders who want to start businesses?

YINGLIAN XIE: I definitely would encourage female engineers and female entrepreneurs who have dreams to go pursue it. For the two of us, even though we were two female cofounders, we went ahead and did it, so I really encourage others to do the same. Gender should not be relevant in this case. When we were looking at this we were not considering that because we are making a technology startup we should have male cofounders, so we shouldn’t do it. It never occurred to us. But I also encourage others to hold the same view to actually move forward and do it.

We two were lucky because we got a lot of support, and I can just hope that the entire community gives more support to other female potential cofounders like, so that they could move forward. This support sometimes comes from family, and without our family’s support we wouldn’t be able to do it. Without the friends’ support and without investor’s support this couldn’t happen. I read Sheryl Sandberg’s book Lean In, before I talked to Fang about our startup, and that also gave me a lot of courage.

ZARA ZHANG: I felt the same way.

YINGLIAN XIE: I feel this kind of experience from mentors and role models could also influence others to pursue their dreams.

ZARA ZHANG: Do you think there are more women leaders in tech in China or in the U.S.

FANG YU: In terms of tech leaders I think it’s probably equal.

YINGLIAN XIE: China actually it feels has less issues around gender. We ran into a lot of senior leaders in China. Because I asked the same question and people ask the same question, and maybe this is an incorrect perception, but I feel like in China this gender bias seems like less of a problem.

ZARA ZHANG: I think people talk about it less. It exists but people just don’t talk about it as much.

YINGLIAN XIE: Yeah. I think some statistics show that the percentage of woman leaders in China is greater.

ZARA ZHANG: Yeah, I’ve seen the same and I think that’s true.

YINGLIAN XIE: I don’t remember the details. I just read some stats.

FANG YU: Maybe when we were growing up in the Chinese education system, men and women were equal, and that was a big influence later in China. I think we have quite a lot of female engineers as well. In terms of hiring we do not consider genders. We have great female engineers and very successful ones.

YINGLIAN XIE: In our company our gender balance is very good.

ZARA ZHANG: Like almost half-half?

FANG YU: I wouldn’t say half, but at least 30% to 40% are females.

ZARA ZHANG: That’s very good by Silicon Valley standards.


HANS TUNG: We are now in the final section of the podcast which is quickfire questions. We will ask the questions very quickly and you just give us the first answer you can think of. Tell us an entrepreneur you admire the most and why? It could be in the U.S. or China.

YINGLIAN XIE: There are many entrepreneurs I actually admire. There are so many great entrepreneurs nowadays if you look at the landscape. I would say very quickly I admire Elon Musk. He’s not just an entrepreneur in one field, he’s a serial entrepreneur of many different companies. And what I like most about Elon is his wild ideas which he pursues really to challenge the boundaries of technology. For the two of us who come from the technology field that is really inspiring.

HANS TUNG: Okay. And you, Fang Yu?

FANG YU: It’s actually very interesting to see Yinglian and I actually have the same pick. That’s why we get along and we started a company together. I read his book, and I really admire that he aims very big, and also puts attention to the details to actually build the things really shine. That’s actually something that we really want to follow.

ZARA ZHANG: Is there something you read recently that you recommend?

YINGLIAN XIE: I read a book called Think Fast and Think Slow. I like that book. It’s talking about how the human brain works and where you develop your conscious decisions. I learned quite a bit out of it, even some concepts that can be applied today in the business running part about, for example, some marketing concepts. I can reflect to that book too and I really recommend it.

FANG YU: I read a not so new book. The Hard Thing About Hard Things. As we said building a startup is a hard journey, but seeing other people in it and knowing we’re not alone is helpful.

ZARA ZHANG: I think that book is most useful after you’ve already encountered some of the things we talked about.

HANS TUNG: That’s right.

ZARA ZHANG: When I read it I knew that I will make the same mistakes, because I just haven’t done it yet.

HANS TUNG: It’s different.

ZARA ZHANG: What do you do for fun?

FANG YU: Currently we don’t have a lot of time, but when I have time I like to travel with my family, spend time with my family to actually go traveling.

HANS TUNG: That makes sense. I like doing that too.

YINGLIAN XIE: I always enjoy reading. I have time where I can read without interruptions, with maybe some tea, that’s a luxury for me.

HANS TUNG: Do you usually read English books, or Chinese books, or both?

YINGLIAN XIE: I read both.

HANS TUNG: What would be an interesting Chinese book for you?

YINGLIAN XIE: I love to read what we talk about is the swordsman book. I don’t know if you know about it.

HANS TUNG: Yeah, of course. “Jin Yong 金庸.”

YINGLIAN XIE: And when I was little I really enjoyed those books, and even today I see there’s new generation, new type of book, and particularly recently there are several new ones about girls. I love those because it’s full of imagination on what a person can do really to pursue the best. I like a lot of the characters. They’re two different worlds. You essentially put the reality into a different word, and then you suddenly feel that some of the things you learned from that world you can take and apply to the reality in thinking a different way, and I interpret that as another book we read before, Game of Thrones, that’s also a popular show now, and the complications there, etc. That’s the part that I enjoy in the books I read.

HANS TUNG: Do you have time to watch any video on Netflix or Amazon?

FANG YU: Not a lot but surprisingly all the videos I watch is with my kids, so kids movies.

HANS TUNG: If you have a little bit of time you give it to them.

FANG YU: Yeah, I watch kids movies and cartoons with them.

HANS TUNG: How about you?

YINGLIAN XIE: Yeah, I try to find the time to do a bit of exercise and at the same time maybe watch something. But what I watch can vary widely from something more serious like documentaries, etc. to just for fun, just purely relaxing content. I see that time as the time I’m winding down every day from work. I need a little bit time to wind down, to exercise, and then afterwards I feel better.

HANS TUNG: When I watch with my wife it’s usually the American ones like The Walking Dead or Silicon Valley.

YINGLIAN XIE: I watch The Walking Dead too.

HANS TUNG: But watching by myself I like to watch more of the Chinese videos like “Sima Yi 司马懿.”So whatever’s the hottest or newest from CCTV or some of the major productions, especially historical ones, I really enjoy watching that.

YINGLIAN XIE: I’ve watched both parts. I’ve watched “Feng Zhen 风筝” which is the really popular one. And I also liked that movie too.

HANS TUNG: It’s fun to be able to watch a variety of shows from different cultures.

ZARA ZHANG: Yinglian and Fang, thank you so much for being here with us.

FANG YU: Thanks for having us.

HANS TUNG: Thank you. It was very fun.

YINGLIAN XIE: Great pleasure.

HANS TUNG: Thanks for listening to this episode of 996. By the way, we also produce a weekly e-mail newsletter in English, also called 996, which has a roundup of the week’s most important happenings in tech in China. Subscribers have told us it is informative and fun to read. The newsletter also features original content and analysis from Zara and me. Subscribe at 996.GGVC.com

ZARA ZHANG: GGV Capital is a multi-stage venture capital firm based in Silicon Valley, Shanghai and Beijing. We have been partnering with leading technology entrepreneurs for the past 18 years, from seed to pre-IPO. With $3.8 billion in capital under management across eight funds, GGV invests in globally minded entrepreneurs in consumer, internet, e-commerce, frontier tech and enterprise. GGV has invested in over 280 companies with 29 IPOs and 22 unicorns. Portfolio companies include Airbnb, Alibaba, Ctrip, Didi Chuxing, Domo, HashiCorp, Hello-Bike, Houzz, Keep, Slack, Square, Wish, Toutiao, Xiaohongshu, YY, and others. Find out more at GGVC.com.

HANS TUNG: If you have any feedback on this podcast or would like to recommend a guest please e-mail us at 996@ggvc.com. This podcast is co-produced by our friend and business partner Kaiser Kuo, the host of the wonderful Sinica podcast. It covers China’s economic, political and cultural issues.