In 2012, Chih-Han Yu co-founded Appier with Joe Su and Winnie Lee in Taiwan. Prior to Appier, Yu spent most of his career studying computer science in academe, earning a bachelor’s degree in the field from National Taiwan University, followed by a master’s degree at Stanford University, and finally, a PhD from Harvard University.

Yu started Appier as an artificial intelligence (AI) company, before the term was even in vogue. Each of its four products help enterprises make better decisions: CrossX Advertising for retention and engagement across platforms, Aixon for predicting customer behavior, and Aiqua for increasing conversions on company assets.

Now a team of almost 400 employees, Appier serves more than 1,000 brands and agencies from 14 offices across Asia Pacific. Appier has raised more than US$162 million, including from Sequioa, Line, Softbank, and Naver.

Ezra Ferraz: What inspired you and your CTO to pursue mobile games as your first venture together?

Chih-Han Yu: My background is in multi-agent artificial intelligence, and so we wanted an AI that could stimulate and a multi-entity that is actually smart. For example, making multi-robots smart or multi-server interactions smart. Our dream was to build a large-scale system that uses AI and that could potentially change the world.

We had the idea of building AI engines for games that could simulate multiplayer interaction. In particular, we wanted the Al players to be able to mimic the real player’s behaviors. We thought it was a cool idea that when you are away, your player can continue interacting with your friends. We’ve basically built a non-stop gaming community. We tried to sell to game companies, but because we didn’t have a background in the gaming industry, that didn’t work out. Instead, we built our own games and properties. Unfortunately, we did not get any experts on board. We are still not very good at that, so our game didn’t work out. That’s why we initially started to build mobile games.

Ferraz: How did your clients let you know they were more interested in your AI capabilities than your game development capabilities?

Yu: Actually, there were a few game publishers that we reached out to after we built our game. They were very nice and gave us feedback on what to change. We worked very hard to improve our engines as well as our games in order to make them work.

In the end, the son of a long-time partner just talked frankly to us. “I didn’t know you guys actually knew AI, because every time you try to deal with AI engines, it’s very difficult. Not a lot of people know how to do it. Why don’t you just take some data from our company and try to see if you can build better gaming engines than our current system?”

What’s interesting is that we used a very simple algorithm that worked twice as well as the solutions they were using at the time. They were very impressed, and we got our first contract to sell. We got a game publishing contract, and we got a data consulting contract. We started to realize the value of our skills and capabilities in this industry.

Ferraz: So, after helping that first client and getting the data consultancy, you decided to abandon your gaming startup?

Yu: Gaming startup, or AI engine for games startup. We didn’t abandon them totally. We just thought maybe we should get more contracts in order to feed our AI engine. After two or three of these kinds of data consulting contracts, we felt that it wasn’t really scalable. People didn’t appreciate our capabilities. Then we decided to pivot to a different type of company that used data and also AI capabilities to solve business problems. We started as a digital gaming company that provided a solution to a software problem. We pivoted heavily to start Appier.

Ferraz: When did you decide to abandon your gaming startup?

Yu: This was back in 2012. That was when a lot of our PCs dominated and disrupted the industry. We have a lot of services that are basically mobile versions of those PC nowadays. It became really popular because people started using mobile phones and spending more time on mobile phones. People became more engaged during this time.

We contacted a group of companies that not only did Internet services, but were also in the customer brands, automobile, and financing industries. They didn’t know how to reach their digital users through these new devices. What was even more interesting is that even when they were trying to figure out the mobile phone, they would say something like, “How do I win the next device?”

Basically, there’s a continuous device disruption. We were thinking that there must be some way to bridge people’s data from one device to another. We were thinking, “How can we solve this problem?” We thought we could use AI to solve cross-device bridging problems. We were amazed and were very excited about it, and we were very good at it, so we solved it quickly and massively and became a very popular solution in the industry.

Ferraz: As more categories of devices came onto the market—say, voice-enabled devices like Amazon’s Echo—did you also add them to your mandate of optimizing advertising cross-device?

Yu: Currently, we remain focused on cross-device as it relates to desktops, smartphones, and tablets. However, it is interesting to track the adoption of such devices and to see how marketers can leverage the data that comes from them.

Ferraz: Can you share how you built your first product? I know one of your concerns was that purely consulting wasn’t scalable, so can you describe your actual first product and how you built it?

Yu: They were two or three data companies, who contacted us if we could do an internal recommendation engine. It would be the best thing for the users, but at the same time, it would cannibalize my current game revenue. The idea was that there were more games, but we don’t want my users to spend too much time on the new games. That was the sort of problem we tried to solve. Then, we built a game engine. It was basically a recommendation engine that could take into account multiple contracts that people wanted to optimize. They could actually optimize the right time value and at the same time, not cannibalize the solo products where we really made money.

We built that first product, and it didn’t work out that well. Only a few companies adopted that solution, and it didn’t become a widely adopted solution. We found that it didn’t have a strong, memorable effect. We decided to pivot away and try to build another product that I think was still a very bad, mechanical product back then.

Ferraz: So what were you doing at this point? Were you leading the development as the chief product development officer? What were you doing as CEO?

Yu: Well, the CEO does everything. [Laughs] Because back then, we were only a team of eight to ten people, who were mostly engineers. I was in charge of product, as you mentioned, and I was also in charge of the big business departments. Some of my co-founders also did this. If the lights didn’t work, I needed to go fix it. I did everything. I can cook, too.

Ferraz: In terms of product development, did you leverage upon your studies at Harvard and Stanford?

Yu: Yes. Basically, a lot of the products we developed in the early days of Appier were all around AI solutions. Even now, a lot of the algorithm background that I’ve developed over the years plays a key role in our product design and our algorithm design.

Ferraz: What did you learn in theory in your graduate studies that you eventually applied in practice through Appier?

Yu: There are a lot of algorithms and concepts we apply at Appier now that source from or are otherwise related to my research studies.

Ferraz: Once you moved past your initial clients, the internal game recommendation engine, how did you sign your additional clients? Did you also participate in business development?

Yu: The game recommendation engine idea didn’t work, and we repeated it a few times. So, we hadn’t really crossed the barriers of bridging technology, which was really getting a lot of traction because a lot of companies wanted to learn how to. I think back then, there was a demand to also start hiring one or two get users through their devices. That helped us grow our business.

Ferraz: What was the target profile of your client companies? Who were you targeting?

Yu: Around 2013, when we were trying out the service officially, we were looking for companies that were PC-based, but wanted to reach into mobile. That was the typical ideal, but it grew into pretty much all kinds of companies that wanted to bridge across devices and channels and gain a cross-device advantage.

Ferraz: In terms of business development, did you have to offer exclusivity? For example, if you had a client in ride-hailing, say Grab, you couldn’t work with Uber? Is it sort of like that, or no?

Yu: It wasn’t like that. They just wanted to see whether it worked. The first batch of clients were just curious. “Can you guys really do that?” We made it work. There was no exclusivity around it.

Ferraz: What was the most common KPI for your clients? Was it downloads or additional purchases?

Yu: Initially, it was only about big capabilities—being able to do stuff. Bridging all of those clients across the barrier was the main KPI. Of course, when our service became full-fledged and was not only all kinds of marketing, but also consumer insight predictions, even consumer purchase and management behaviors, it got more complicated. Each product had very different KPIs. Initially, the KPI was very simple, and then we brought it to that, and people were very happy about it.

Ferraz: Were all of your clients primarily based in Asia and Southeast Asia, or did you also service clients from, like say, Silicon Valley?

Yu: Back then, it started in Taiwan, and then Singapore. Only two countries.

Ferraz: Why did you settle on your home country, and why Singapore?

Yu: When we were trying to do some research around Asia. We bootstrapped ourselves, and that was actually interesting. Basically, the company didn’t make any revenue for two to three years, and then suddenly, three months after launching the service, we were able to make a profit. Not only revenue, but profit. We were very excited about it.

We were counting our profit, and we thought, “We can hire one or two international salespeople.” We started thinking about where to put these people. In Japan, China, or in Singapore? We started going around these countries and also talking to salespeople with industry background who wanted to join us. It turned out that the first person who committed to join us was from Singapore, so we decided to start there.

At the same time, the media was releasing reports that China and Japan were the two biggest economies in Asia. The reason we chose to explore Singapore was because it was in the top four in Southeast Asia. We didn’t think we’d grow really fast in those few years, so that’s why we chose these three places to start exploring.

Ferraz: Out of those initial markets that you had, can you share maybe one of your success stories on how you helped one of your clients?

Yu: There was one really big department store in Taipei. It had some Internet traffic and people, in the past, always went to their website to check their discounts, but people started spending more time on mobile so the store sharpened its approach to those digital customers to show them discount messages when they were around their department store areas. It still sounds impressive now, but it was super impressive back then because people wanted to bridge devices, but we also did location-based intelligence. That was a type of cool technology to get to. They were super excited about it.

Ferraz: What was the positive outcome for that client?

Yu: The outcome was that we were able to get people to carry their mobile phones when they walked through the store. They used to be interested in their service and then the department stores. Because of our solution, we were able to bring them to actually visit the department store.

Ferraz: What inspired you to pursue your second product, Aixon? How did you see this complimenting your original solution?

Yu: That was very natural. Originally, our motivation was in large-scale data and our AI capability to solve business problems. That has always been our vision and also our motivation. Our product across the marketing area allowed companies … once we established long-time partnerships, to say to us, “Can we give you our data for you to build our in-house capability? Because you can help us predict what kind of customers are interested in my product and what kind of people would actually be looking for our services.”

We internalized show them in-house capabilities. It started two years ago. We had a goal. We hoped that we would be able to empower our customers to build their automotive data science team in-house without the hassle of building the whole team there. We started to build a data science team on top of the empowering capability of being able to predict consumer behavior. That was sort of the motivation behind building our second product, Aixon.

Ferraz: Can you give a few examples of what type of consumer behavior Aixon can predict? I understand it gets better over time, so can you also share about that?

Yu: Actually, it’s super powerful. You can predict and define your own goals. For example, if you want to predict who would be interested in certain products, who is exposed to certain products, or you can define these and predict whatever goal you want. That’s the first capability.

The second capability is that it can let you understand your own gains, not only on your web store, but what kinds of topics customers are interested in on your website—down to an individual basis. We also provide data and prediction capabilities for outside your website.

So you can imagine, someone opens the door and you can not only observe anyone inside your store, but you can actually know what happened outside of the store. Pretty much, before the customer even comes in, we already know what to sell to them. That’s how we empower or allow our customers to have super capabilities like these. It’s really driving the difference. They can define their own goals and are able to know their territory. They get insights on behavior from outside and inside. It’s the most powerful capability that Aixon has.

The third one is being able to segment users. Every user and every customer who visits you or becomes a customer is different. How do we smartly segment them and have very personalized media conversations? That sets apart Aixon’s capability.

Ferraz: Can you share a story of how Aixon has helped a company over the past two years?

Yu: We not only helped a company with better recommendations for their customers, but also generated more revenue. One top media group in Taiwan has digital media that’s on the Internet. It’s a bit related. They also do a lot of activities and try to re-engage with the people who are interested in the business. They have a lot of different offerings outside, like magazines and online media. They also have regular courses for executive training. One of their biggest problems is that they don’t know how to correlate their current database users with their more personalized, in-depth offerings.

Those in-depth offerings may be executive trainings, or maybe they are financial products we help them with the user’s topic of interest on the website and also onsite behavior—we can collect a very holistic view of their audience’s interests in financial products and financial topics.

They can segment the audience into very fine commonalities and very personal conversations with promotional messages that can be sent to them. They are very grateful for the service provided by this tool.

Secondly, because they have very tremendous data, we’ve also helped them segment those data and promote to people with all kinds of financial needs. To reuse the data, monetize their audience, and debunk group companies that sell competing products.

Ferraz: I mean for Aixon, and just for Appier in general, you have to have a lot of technical talent. You mentioned having a data science team, and I know you’ve been really successful at recruiting top talent not just in Asia, but from around the world, including Silicon Valley. As a founder, why do you think you’ve been so successful at recruiting top technical talent?

Yu: Top technical talent, loves to work with other top technical talents. We really want to build a community and atmosphere where people really appreciate each other. That’s the most important thing. We have been fortunate because my co-founder is very connected. He was part of a national team for programming in Taiwan.

When we started, we only had a few people, but my two co-founders national programming team members. They represented the country. Taiwan actually is probably one of the top-five or top-ten most competitive programming national teams those were very interested in the topics they are pursuing. They created a very positive recruiting look. That’s how we started from a few talents to a group of great talents.

Ferraz: I see. So, other than trying to bring the best talent—because great talents like to work with other top people—in terms of culture or benefits, how do you make it very developer-friendly to help attract these talents?

Yu: I think programmers usually want to work in challenging departments. We’re always trying to challenge ourselves to do what’s next. We have one solution to a problem, but we immediately jumped into developing our next generation solution and finding a problem to disrupt. Taking on challenges and being able to address these problems excites people. We also ensure that they have their freedom and that they have some autonomy to choose the way they solve problems.

We also encourage them to use new tools and technologies. We don’t use a very conservative approach when developing a new product. Whenever they hear there’s a great tool or hear about someone publishing a powerful algorithm, we always encourage them to try. That kind of culture and mentality really is appreciated by a lot of the great talent at Appier.

Ferraz: Are you still very actively involved in the recruiting side of your top-level talent?

Yu: After we’re done here, I have to finish a whole series of recruiting interviews. [Laughs]

Ferraz: My next question is more related to the business development and sales side. You generally service businesses in three categories: consumer brands, e-commerce companies, and mobile commerce or game developers. Each of these has its own unique sale cycles and idiosyncrasies. What are the challenges of selling Appier’s solution to each of these three categories, and how do you deal with it?

Yu: Each category has different needs, and what we do is to really understand the customer’s needs and provide the appropriate solution. Basically, when should we present X model or Y model to match the customer’s needs? That’s something we need to provide in our internal training. We have been actively dedicated to training our people to be able to match the problems that the customers face with the solutions that we are able to provide. We still think that one of the best things about us is our customers. To match the solution that can solve their most urgent needs is the most important thing.

Ferraz: You mentioned just a few moments ago that what attracts developers is solving big problems. What are the next big problems that you’re looking to solve?

Yu: I said from the get-go that AI was going to change a lot of things that we do nowadays. There’s one saying recently that I think really matches the current situation. “There’s no standalone AI industry, but every industry is transformed by AI.” Basically, we will continue. We have seen a lot of opportunity in each industry. For a specific push that we can solve problems, or there are processes we can automate or make more optimized—those are the opportunities for AI. We think that there is a lot that can be foreseen—some of them are relatively dead already, some are bigger markets.

Ferraz: You now have a well-funded war chest of venture capital and serve more than 1,000 clients worldwide. Of course, Appier is successful, but how do you define Appier’s 2,000? What is your ultimate goal for the company?

Yu: I think our success doesn’t rely on whether we can deliver on the very best solution and on how we drive change for the customers or the industries they are working in. Our success is not defined by the amount of funding we’ve raised. It’s also not defined by the number of people we have. It’s rather defined by the success of our customers, by how much we have changed the way businesses connect and interact with audiences.

Ferraz: I understand you have customers all over the world, including some of the top tech companies in Silicon Valley. What are the challenges of being an Asian company servicing Silicon Valley companies?

Yu: If you really have a good solution, there are no particular challenges. Maybe sometimes there are some differences, such as we cannot have real-time communication. That’s pretty much the only one. But we think that now, society really has no boundaries. There’s no specific challenge.

Ferraz: How do you see AI changing the lives of consumers in ten to fifteen years? Just the everyday person, not a brand.

Yu: I think you can imagine everything will become intelligent from the most complicated things to the most interesting things. I remember when I was doing my PhD thesis, there were these body chairs that can ultimately save people’s postures and provide the best support for the lower hip. I think someday, that day will come. Starting from the most immediate technologies to passive devices and furniture and the Internet of Things—all will become intelligent. We will be surrounded by intelligent devices and be making decisions in a much more abstract way. That’s what I perceive in the next ten to twenty years.

Ferraz: Even today, looking toward that future, how can even small businesses leverage or get into artificial intelligence if they’re not the size of Gojek or Grab?

Yu: The first step is to start collecting the data. Organizing your data well is very important. Data, in the future, will become a very important asset for any business. Not only for big businesses, but also for small businesses. Small businesses will also grow bigger. In the process, an organized way of collecting data and records. Media across time, across different customers, or across different services and products that will actually inform you to make much more intelligent decisions in the future.

Ferraz: Entrepreneurs in Asia get all sorts of advice from the media or from other founders. From your perspective as the CEO of one of Asia’s top startups, what advice do you give to people who want to build a company as successful as Appier?

Yu: I wouldn’t call us already successful, because we’re still in the process of building a successful company. I think it makes no difference building a great company anywhere in the world. You need great talent, great ideas, and great execution. It’s all about that.

In Asia, I think there’s a great opportunity, because the economy is evolving at a really fast pace. How do we localize a lot of our great ideas into each market to capture that opportunity? That has become a very important subject.

In Asia, it’s probably that the community of entrepreneurs isn’t as active as in Silicon Valley. Here, you be more independent. It’s unlike in San Francisco, in Silicon Valley, where if you’re working on something, you can ask people for advice, and they’re sitting right next to you. Here, having a much stronger mentality is very important.

Ferraz: Because things are different here in Asia as compared to in Silicon Valley, how should founders be more proactive at finding great talent, learning how to execute really well, or learning how to find the best ideas? What advice do you have given our cultural context?

Yu: I think there is no fundamental difference because our community is more spread out. Being able to sort of excite yourself and always pursue a better way to work is important. What other people do may not be what you should also do. You need to have your own determination and your own mentality.

Amongst all my friends, they have been doing very well in Asia. They are probably the first companies of their category in their countries. A lot of the time, you can learn from experience from other categories, other countries, but sometimes you still need to make your own decisions based on what’s best for the company and what it’s going to attract. If there’s no one to guide you, how do you find your way around that? Those things become really important.

Ferraz: You mentioned that a lot of your friends are the first companies in that category. What advice do you have for founders who want to start a business, but don’t know yet what problem to solve or product to build? How do they find their great idea?

Yu: I think great ideas need to be validated by markets. It’s very hard to hit the nail on the head, so basically, you have to find the problem that people currently have that you believe you can solve, but at the same time, you also need to consider multiple constraints, such as time, money, and regional differences all together—whether those can be solved with your current capabilities. If not, then how can you bootstrap and eventually build toward success to achieve your goal? That’s the sort of mentality you need to be putting into place.