Cross-screen targeting based on AI marketing data analysis to map the user journey, generate real-time predictions, improve attribution and increase conversions? It might sound like marketing nirvana but Appier’s CEO Chih-Han Yu is making it the new reality.
Marketers are always chasing the next breakthrough in marketing attribution.
From the largely outdated single-source model of user attribution to more elaborate probabilistic algorithms, the evolution of data-driven marketing is inseparable from the evolution of our understanding of attribution.
Never before has the demand for attribution been so great. This comes as businesses invest in increasingly disparate, omnichannel marketing campaigns guiding customers along a sales funnel that looks more like two brawling octopi than the traditional inverted pyramid.
Chih-Han Yu, Appier’s founder and CEO, had an epiphany chasing the answer to attribution from an artificial intelligence perspective. He realised he’d never find it unless he focused on business problems first and let the technology flow on from his customers’ specific demands.
And the most common need he found in the marketplace? Cross-device analytics.
“Using our platform, [for example] we can attribute that the user always checks the product on mobile device on their way to work,” Chih-Han says. “Then when they go to the office the users start purchasing on their office desktop. So we can actually link and complete a story of a user journey, across devices.”
Appier is a SaaS platform which draws on artificial intelligence to track single users across multiple devices – and even provide real-time predictive analytics for any given user based on the previous buying habits, to drive personalised marketing campaigns.
He gives the example of users who always window shop on their mobile device as they commute – but only makes purchases on their desktop at home. By deploying broader ads to mobile users and personalised desktop marketing to that user on the way home, Chih-Han Yu was able to assist the company increase conversions.
Tune in and join hosts Nicole Manktelow and Mark Jones in this deep-dive episode of The CMO Show discussion as they talk marketing analytics, attribution, startup culture and of course, Boston Dynamics’ Big Dog.
The CMO Show production team
Got an idea for an upcoming episode or want to be a guest on The CMO Show? We’d love to hear from you: firstname.lastname@example.org.
Hosts: Mark Jones (MJ)
Nicole Manktelow (MW)
Guest: Chih-Han Yu (CHY)
NM: Welcome to the CMO Show. I’m your host Nicole Manktelow.
MJ: Yes, and I’m Mark Jones. We have a very special guest today. Chih-Han Yu is the CEO of Appier. Thanks for joining us.
CHY: Thank you for having me.
MJ: First question. What is Appier?
CHY: Appier is a company that have developed an artificial intelligence platform that help in the process of solving various business challenges. For example, we’re helping to be more precise in marketing or better in analysing and understanding their audience. We are formed by a bunch of passionate and experienced artificial intelligence scientists, as well as marketers and also data analysts from different domains.
NM: I’m trying to work out what a bunch of passionate, artificial intelligence scientists get up to when you leave them all alone in the room. What have you guys been working on?
CHY: Yeah, I think it’s very important that we don’t only focus on technology or algorithms that we’re developing in AI, but we also focus on the business problems we solve. It’s very important that we bring the demands and also the analysts and you know, what the challenges and the problems that our customer is facing.
MJ: Right. Before we go much further, artificial intelligence is all the rage, but how are you different?
CHY: We have been operating for five years. Across these five years we have profiled and understood how to build devices across the Asia, Pacific region. Pretty much 80 percent of the internet population we, anonymously we know what device they own, and how to use different devices. What kind of context and what kinds of items they intend to purchase on which devices. With this information we help business to understand the user much better.
NM: How many user profiles do you have?
CHY: Two billion.
MJ: Two billion user profiles?
MJ: It’s worth explaining. Correct me if I’m wrong but what your technology does is it allows you to develop a user profile based on multiple devices. So one person, anonymously I presume, you can say they have, let’s say, an iPad, a smart-phone of some kind, a PC, a laptop or something. Maybe three, four, or five devices. You can say “those devices are all used by one person.” That’s part of your technology, isn’t it?
CHY: That’s exactly right. For the first act we profile and also figure out what we call device connectivity around identity. The second step, then we also further infer, in real time how people are using each of the devices, and what is their intent or interest on each of the devices.
MJ: One of the biggest challenges I can see in all of that, aside from the privacy question, is how do you do that when there’s such a rise of ad blockers and other technologies? People are masking where they’re coming from. It’s become a really big issue.
CHY: Yeah. For those people, there’s some of them, they have ad blockers on all the devices and we might not be able to track them. But sometimes we still can attempt… One very important thing, we don’t use any data with the PIID information, we only use 100 percent anonymous data.
CHY: So there’s no linkage with their real identity. So safely we process those data with anonymous identifiers. Some of them, if they have ad blocker in every device we might not be able to track it, but from our experience, it’s only a very small portion of the population, that we can’t track.
MJ: What percentage? I’m interested. Is it growing?
CHY: Because we operate across so many different countries, and each country has very different percentage, I would say single-digits percentage, on average across Asia.
CHY: We don’t see a significant growth. Sometimes when you look at it from the marketing or advertising point of view, they just pre-filter. Those people don’t want to see ads so those will help you to do more precise targeting by layer.
NM: Is it tricky doing this kind of business across Asia, where the rules in the countries are very different, particularly about privacy and what the different governments expect companies to be able to share with the data?
CHY: Yeah. The way how we operate, in order to overcome these cross-country regulation differences, what we do is we use a more strict standard in our database. We don’t collect any of the PIID information.
CHY: Personal Identifier ID.
MJ: Right. Okay.
MJ: So presumably you have the equivalent of an avatar or you create, or a number or some way of assigning a person to a number or whatever you do, right?
MJ: Right. And then you go from there.
CHY: Yeah. There’s no way we can infer back to the identity. That’s the most strict standard that we can apply across the countries.
MJ: So in the homework that I did, in reading up about your business, there’s two aspects of it. One is the data collection side. You’re collecting all this data. Secondly you’ve got a platform which you sell to marketers and business people, right? So people can come to you and say “I want to get access to a certain demographic”. Presumably you’ve got the ability to let people buy from certain demographics, whether it’s age or gender and so-forth, in certain locations, is that right? Is that the two sides of your business? The data collection and then an access platform?
CHY: Maybe I can explain a little bit. One side is we collect data and also infer a complete profile of the user. What devices anonymously and also how they use each device. On top of that, then we provide various targeting capability. For example, I want to target this person on a certain device that they are more likely to buy sports product, or more likely to subscribe to some business. That’s what we call CrossX, which is a programmatic platform that is leveraged on top of our two billion across-device profiles.
We have a second product which is enterprise data intelligence platform. That’s called Aixon, called AI in actions.
MJ: Okay. Sorry, how do you say that?
CHY: So it’s awesome, right, put AI into action. Because currently, a lot of businesses, they have various types of data. Actually they are associated with different ID from different devices. But we have the ability to help them to unify, to connect all that data together. So you have a holistic view about their customer/user across the data collected from departments at different sources.
CHY: And we provide a platform that can help them to condense that data. It’s a SAAS platform, cloud-based SAAS platform. The second stage, after the data has been linked, we also help them to choose, because normally people want to use, want to predict and want to infer what kind of items that my customer wants to buy, and who would try and find my service. A lot of questions would want to forecast, and those are to infer the future behaviour of our customer.
MJ: So it’s predictive analytics.
CHY: Predictive analytics.
MJ: Attribution is another way of looking at this whole challenge that you’re addressing. So marketers are looking to attribute spend on marketing to dollar outcomes, so return on investment, sales, et cetera. How are you helping them solve that, in an explicit way?
CHY: So we are actually solving these problems with a very significant contribution. Usually one of the bigger challenges for a digital marketer is “I have a lot of people, a lot of users purchasing on one device, but how come I have a lot of traffic from the other device but the user never purchases?” I’ll give you one very real example. I have one e-commerce customer. They’ve been wondering “Why do I have a lot of customers viewing my products on apps, but no one actually transacts on mobile devices, but always transacts on desktop devices. Whether I should increase my spending on desktop device versus on my mobile device.” But using our platform, we can attribute… clearly see that the user always checks the product on mobile device when they’re at work, and on their way to work. And then when they go to the office, I don’t know why, the users start purchasing on their office desktop. So we can actually link and complete a story of a user journey, across devices.
MJ: So you’ll get a better ROI for putting your dollars in mobile advertising, or in that dynamic that you spoke about.
CHY: Yes and then what we did for that customer is we enhanced the digital advertising to increase the item recommendation on mobile device. So they use the new mobile device. So we figure out their interests and intent more clearly. On desktop device they would push more promotional and transactional device to enhance the conversion.
NM: So it’s your impulse spend when they come into the checkout… Put that right in front of them when they open up their browser.
CHY: Yes. And then that dramatically increase the conversion rate by at least 30 to 75 if I remember correctly.
NM: They’re big numbers.
MJ: Well absolutely, that’s significant.
MJ: You’ve raised something on the order of 30 million dollars off the back of this idea, right? And the technology that you’ve built, you must have explained this about 100, 200 times right? To all of these investors, but how do you take an idea that you’ve just described which, by the way, that idea that you’re talking about is a very big idea in marketing. It’s almost like nirvana. If we could know these people now but also in the future and we can anonymously present ads to them in the right context when they’re ready to buy. That’s a big idea that a lot of people are chasing, least of which is Google and Facebook.
How have you turned that idea into this fast-growing business? Because I understand, from your background, you’ve got this great research pedigree of Harvard and Stanford and so-forth. How do you go from that engineering, idea-generating perspective into millions in funding, offices around the world, growing a business. You’ve got a lot going on there.
CHY: Yeah. We have been lucky. We have raised so far 50 million dollars, including a lot of world-renown institutes, including the top investors such as Sequoia Capital, Temasek Holdings, and a lot of great investors. But that actually came last. The first is we understand our customer’s problems and demands. The second step is we invent a technology to solve those challenges/problems. The funding actually come last. After we solve it successfully, funding actually is just a proof-point that our team can execute to scale this solution across different investors and different countries.
NM: I’m just curious about the invention process. So you know you’ve got the smarts and you’ve identified the business problem. “Look, I’ll just invent the solution and then the investors will come,” but I’m sure that invention process was quite involved.
CHY: Yeah, yeah. Maybe I can tell a little bit of detail, a story of invention. Just like a lot of artificial intelligence researchers, I was quite intrigued and also I quite believed in the algorithm that you invent and I thought that was the best of all time. I really wanted to make my algorithm work and get used by a lot of people. So I think back five, six years ago, I started applying to think “How can I apply my algorithm into different fields?” But that didn’t work out. I tried multiple times. Our team have failed in eight products or prototypes before we have got one idea really scaled.
CHY: After we failed a couple times, then we changed our mindset totally. Market first. Then technology second.
MJ: The classic pivot.
CHY: Yep. Classic pivot. And also we tried to pivot fast. It’s also an art. Sometimes you can measure by size whether you should persist on one idea, whether it’s maybe just six months too early, you should just persist a little bit. Or sometimes you should pivot completely. So far that kind of decision is more art than size. But we have been through a couple of times.
MJ: Or gut! Gut feel, right?
CHY: Gut feeling.
NM: It’s exciting, isn’t it? I can see it on your face. The art is exciting.
CHY: Yeah. That’s a part that always…
MJ: So is that what you mean by luck? Make the right gut feel in terms of which way to pivot and take the organisation?
CHY: I would say it’s partially luck but a lot of times it’s a lot of analysis and a lot of strategy that comes into place to make the right decision. [0:14:54] MJ: Aside from any kind of exit strategy that I’m sure your investors are looking for, what are they most interested in, what’s the thing that investors have been excited about?
CHY: If you look at current artificial intelligence landscape, world-wide, 60 to 70 percent of the leading AI companies are concentrated in the US. Another 20 percent are concentrated in Europe. Another five to 10 percent concentrated in China. But you look at Asia Pacific region, including Australia, and New Zealand, there’s no leading AI company. But if you look at the projection of AI impact business-wise, economically-wise, actually this region will be the biggest, world-wide. But there’s no single leading artificial intelligence company. Early this year we were selected by Fortune Magazine as a top 50 AI company, and we were the only company selected in this region.
MJ: Okay. Congratulations!
CHY: Thank you. We have been lucky.
CHY: Yes, 2016 personally, myself was selected as someone as probably the first AI person in the region to be selected into the AI forum. And this actually represents a big tremendous opportunity in the region that if we have good AI technology, that we can solve. We can actually turn into a very big business impact. Our company actually gathered a lot of top artificial intelligence scientists to work together to solve these challenges and big problems.
MJ: So AI, yes you could say the AI industry is worth X billion dollars, but really it’s an enabler for something else.
CHY: As in what you say, that is correct. AI is about being able to solve multi fields. So many different fields in the business question. AI eventually needs to bring the economic and real business impact to the market. We’re not talking about beautiful AI algorithm, beautiful mathematics, we’re talking about really deliver satisfactory results to our customer.
CHY: So that’s the most important thing. That’s AI, what it should be heading to. For the question about job security, I think AI will be not just enabler for certain fields, actually an enabler for our life. A lot of times with various things we do, you will be human beings’ capability to do things faster. If you look at nowadays information flow, we simply have too much information on a daily basis. Imagine there’s now a thing called filter, an intelligent filter or intelligent recommendation that filters the most-important information that you should absorb in your life. Your capacity simply the biggest limitation of your work capability, and AI will come to play.
MJ: It’s really been awesome to understand where you’re coming from. Can you give us some examples of what you’ve been doing? Some customer case studies?
CHY: Yes, sure. We have worked with various different industries and customers. Many of our customers come from our brands, e-commerce, and also mobile apps such as O2O and also mobile commerce spaces. For example, one cosmetic brand, they will use us.
CHY: Yep. They have their retail store, they have online shopping, they also have e-commerce sites. What we do is we help them to integrate the data that they collect from various sources altogether, and then connect by using our personalised profile to link their customer together and then we analyse how their customer journey is like. Then we’ll use marketing intelligence solution to enhance the best possible conversion funnels across their different experiences. The second step after we have enough customer data, then we use Aixon platform to help them to predict, and also analyse the customer’s future behaviour.
So that’s usually three phases how we work with our customer.
MJ: How long does that process take?
CHY: Usually one or two months, because we need to have some time to do data synchronisation.
NM: How much data do you need on a person to get a profile where you can predict what they might want to buy, and what timeframes?
CHY: That’s a great question, because some people are more predictable. They have certain patterns. In that case, fewer data points might be sufficient. But sometimes people have very complex behaviours. I would say the more data is better. The general rules that we have, we need to observe and collect people for a few weeks of data to find a reason and also infer the activity of a user across devices.
NM: Do you think this kind of technology favours bigger-ticket items or could it be used in the fast-moving goods section?
CHY: Of course. I think fast-moving goods because for example, in our platform, we not only do a profile of the long-term behaviour, we also do the real-time and on-the-fly inference to complement that part. Because a lot of time people’s decision is quite instantaneous. You need to catch at a moment that they’re more likely to purchase and make a purchase decision. So that kind of decision you need to be very quick, very time-sensitive.
MJ: That’s been one of the biggest challenges, is the idea that the data is actually like fruit, for example, it goes off quickly, right, it has a shelf life. So you’re saying it’s real-time. We kind of stay on it.
CHY: Yeah. Because the real-time behaviour is also a very important component to predict what someone would do for the next seconds.
So, what is keeping you awake at night. It sounds theoretically like you’ve solved a lot of these big problems but clearly you have a long way to go. What’s the biggest problem that you’re facing right now?
CHY: Whether we can solve those biggest challenges fast enough. I think that’s usually our biggest challenge. And secondly, whether we can provide a solution really fit to each market, each customer’s needs. That’s always the thing that keeps me up at night. I think, although our company’s innovating at a fast pace, but never fast enough.
NM: I want to ask you one thing about AI. What drew you to this line of work?
What appealed to you, when you were a student, what grabbed your attention?
CHY: I think it was 16 or 17 years ago, and I was a computer science student, undergraduate student. This one time, I still remember, I took a class called Pattern Recognition.
MJ: Pattern recognition. Okay.
CHY: There was a project that is we need to build face recognition. Which now is a very common application for artificial intelligence, but back then it was still a very new field.
I felt I could stay up all night just to make a face recognition algorithm work. I was so intrigued with, and also I was so interested in AI then, because I’m Taiwanese citizen, I need to serve in the military. Compulsory military service for two years. I still remember when I was serving in the military, whenever I have time, other people would try to do chat and gossip. I always get some research paper and read about face recognition technology, because I wanted to keep myself updated about the field, outside.
MJ: Right. It sounds like that was a long two years for you.
CHY: Yeah. And I was fortunate to work in Stanford artificial intelligence lab, also experience after finishing my military service
NM: I’m a very big fan of them because of the robotics and the videos of the robots doing crazy things, so I’m shamelessly… I rather love them. Did you work on the robots?
CHY: Yes. I think back in 2003, I started working on a project called Little Dog to make a four-legged robot being able to learn to stand, and then being able to walk, and then negotiate across multiple different terrains, purely by learning, instead of by pre-programming strategies. I collaborated with Boston Dynamic. Back then they produced and also helped to build this amazing robot called Little Dog, and also the bigger version called Big Dog.
MJ: Big Dog, yeah.
NM: I think I’ve seen that.
MJ: I think that’s the one I’ve seen being kicked over. Is that the video where some guy kicks it and it stumbles and then stands up again?
CHY: I applied my algorithm, on a junior version, sort of a broader version of Big Dog, back in 2003/2004. That was a DARPA project.
MJ: That’s great. Well, all the best with your company and your team. I think you’re onto something really big. It’s quite clear that there’s lots of work to be done, but it’s very exciting to imagine how we can start solving these attribution challenges, return on investment, all those big marketing challenges, but also the technology side of things is really quite fascinating. Before we let you go though, on our show we like to do what we call rapid-fire questions at the end.
MJ: And this is a chance where we can get to know you as a person, a little bit more.
CHY: That’s very exciting, that’s new to me.
MJ: What are you grateful for?
CHY: For the opportunity to work on exciting the problem that can always excite me.
MJ: Which do you like better, the beach or the mountain?
CHY: The beach.
NM: Who is your hero?
CHY: Steve Jobs.
MJ: I thought you were going to say that. If you weren’t doing what you’re doing right now, working in marketing and AI and solving these problems, what would you be doing?
CHY: I’d be teaching at a university.
MJ: What did you eat for breakfast today?
NM: What would you have rather had?
CHY: Rather have some meat!
MJ: What was the last conversation you had with your parents?
CHY: It was yesterday, I tell them that I’m about to depart to Sydney.
MJ: If you could change one thing about marketing, what would it be.
CHY: I hope we can unify a standard of data collection, so data can be stored safely and also easily.
MJ: That’s a very good response.
MJ: If you had to change your first name, what would you change it to?
CHY: My first name? I wouldn’t want to change – some people call me “CH” Maybe we could make it shorter, changing it to CH.
MJ: There we go. Well CH, thank you for being our guest today. All the best with Appier and having an amazing time flying around the globe and changing the globe at the same time. We wish you all the best, thank you.
CHY: Thank you so much. Thank you so much for having me.