Episode 98: Customer Analytics in the Age of Machine Learning

Jon Prial: Everyone knows the story of the mythical founder, right? The wunderkind who's got a great idea and winds up turning that idea into a successful company. Today, we're lucky to be talking to one such person. Derek Wang who brought his ideas from academia, is the founder and CEO of Stratifyd, a company with an AI powered end to end customer analytics platform. Stratifyd is one of our recent investments and on the show we'll be talking about Derek's differentiating vision and now he and his colleagues have taken a trove of unstructured data and applied it to a specific market space using Machine learning and NLP to deliver valuable insights. We've got a very interesting episode ahead on the technology Derek uses to capture end user interactions of all kinds and then to provide rich, actionable intelligence to his customers. Stick around. I'm Jon Prial and welcome to the Georgian Impact podcast. So, welcome Derek.

Derek Wang: Thanks Jon. Thanks for having me.

Jon Prial: Our pleasure. So, do me a favor. I'd love for you to take us through kind of your background and kind of a little bit about this transition from academia to ending up at a startup.

Derek Wang: Yes, glad to. So, before I started Stratifyd, I was a director of the Charlotte Visualization Center where we do a lot of applied research scientists with the US government's related agencies to kind of helping them to identify what are the emerging topics out of all the unstructured data they have. So, part of my motivation jumping from academic to commercial side was actually my experience of personal education, when I was working with Microsoft Research and Xerox PARC, during my PhD time, that they're kind of the combination between apply size and business cases really is one of the motivators are looking at it.

Jon Prial: That's great. While you're getting your PhD, you're actually being to get a taste of business. So that's kind of... I like that. Maybe everybody should do that. I think that's a great way to kind of break the barriers here.

Derek Wang: As an applied scientist, yeah. You got the chance to work with the business units, understanding what their problem is. And that goes to the second part where motivated me to start Stratifyd is doing my academic years, I got the privilege to work with a lot of bank business units. So really gave me the chance to interview them and understand what is their true unstructured textual analysis problems that they're trying to solve, or is there any a platform that outstanding that they're using. And the answer to me is no, there's still a lot of space that need improvements.

Jon Prial: Interesting. So you actually started with unstructured kind of with the three- letter agencies in the images, and even the banks are unstructured. And as we get later in this podcast, I want to start talking about how you begin to marry that with some of the structured data as well. I will get to that in a bit, just as you began to start your company kind of get out of academia and then begin to start the company. What were some of the challenges that you faced in terms of building a team and getting the skills required? Because obviously this would be relatively new for you.

Derek Wang: Yeah, absolutely. As always people were saying, Hey, when you beginning the CEO is actually wears a lot of C hats, like chief janitor officers and all that. So I started with two of my co- founders one was my PhD student, one was my post- doc three very technical person where I want them to focus on attack where I can start branching out on the business side commercial. So a lot of things are figuring out that the registration, the legal, the financial and everything. So that's a little bit tactically a challenge, but more challenging in running the company in that is now you start really need to build out a mission, right? You're not just delivering a product, you're building a relationship, customer relationship. You're building your sales side, which is a lot, which fortunately, I like to sell stuff. I like to sell our company, our product and talk to people but I think a lot of the things that I didn't anticipate is how much of a people business attack startup is really is. So a lot of that is what I learned.

Jon Prial: But the question that follows up in terms of skills, because we talked about the people you brought with you. So you're in North Carolina. What challenges did you have or were there any challenges in growing skills at North Carolina? Because it's good to see this world of AI being not just anchored in the West coast or a few small cities. So how did that work for you?

Derek Wang: Yeah, I think frankly, we haven't noticed any challenges in terms of talent hiring. Part of the thing is we basing in Charlotte, which is a very, a bank city, a large inaudible force in the bank, a lot of business units in this territory, but also with the way I'm looking at talent acquisition and all that, it's all about culture, right? So Charlotte has what I always say, Charlotte has become very hip. So in the past three years, we got at least 10 different breweries, like microbreweries opened up in the city. And it really, the other piece of this is the quality of life, right? So when you go to bigger city, even West Coast and Seattle, those are highly stressful high pay jobs, but also on the flip side, it's very expensive to live there or what we can offer in North Carolina and the city, it's really a well balanced work and life quality, and then people will feel a good supportive community here.

Jon Prial: That's great. So one last question about the role as a CEO. Did you feel comfortable finding the connections necessary as you went through some of your early financing rounds?

Derek Wang: Well, that's a really a good question. None was in North Carolina. I think North Carolina it's considerably drier than in terms of financing landscape. So we were fortunate the first round of our money was raised out in Silicon Valley. Although we got a lot of push back of saying, you need to kind of move to Silicon Valley. Financing side it's a little bit challenging, and we're just fortunate that Georgian partner out in Toronto, we will be able to make that relationship and connection, but generally Charlotte region, North Carolina, is a little bit tougher around on that part.

Jon Prial: inaudible Trade- offs. So let's talk about Stratifyd a bit. So, voice of the customer, customer analytics, these are pretty well known markets that we don't need to go into, but I'd love to hear for you what makes Stratifyd different in this space?

Derek Wang: Yeah, so we started by looking at unstructured data, which is 80% what where your customer interaction actually state so. So unstructured data is the messy part where you're looking at customer comments, interactions, chats, and etc. So we use machine learning and natural language understanding to unlock all those hidden topics for you in near real time. And the real time part is second differentiation points of our solution, which every company knows those customer interaction data are perishable. So there's a time window, how fast you will want to lock those insights out of those data. And we are here to help you unlock it in near real time.

Jon Prial: Interesting. Now, as I think about this a little more and the nature of the learning since you're doing real time and you move very quickly, is there an element of how you train the system or is this kind of all unsupervised learning? How do you begin to begin to pull all this stuff together? Because a call center from an airline might be very different than something you might learn from a comment on a website about a hotel, for example, how did you kind of pull the technology together under this?

Derek Wang: Yeah, so that's a great question. So our design principle actually goes back to the human center, right? So number one, we are a big gain. As we trademark augmented intelligence is really, we use AI to augment how the business unit can perform their analysis. And when we started the learning part, to your point, is we leveraging what we call unsupervised machine learning, to start the process. So essentially we don't need to manually code airline data points versus a tobacco company, just for the sake of argument. But we leverage our AI to automatically read through every single comment or words out of your costumer interaction and dynamically generate those topics in the unsupervised fashion. So is purely data driven. You'll get that. And then we visually represent all of those insights to the decision maker where they can start kind of using their human intelligence coupled with those machine intelligence to making decisions or identify patterns where then they can start taking that and plugging into our supervised learning, deep learning neural network module of the platform, generating automated workflows and decision making process. So it's really humane the center, unsupervised process first, and then supervised machine learning to kind of automate the whole efforts.

Jon Prial: That really completes and defines kind of the feedback loop. So that really-

Derek Wang: Absolutely.

Jon Prial: So obviously we couldn't have done this 10 years ago. We didn't have the ML. We didn't have the deep learning necessarily behind it. When you look at some of the techniques you're providing, is it more about as you get to this learning and the labeling that the end- users are doing, where does something like maybe sentiment analysis show up, for example.

Derek Wang: Yeah. So really when we're talking about this unsupervised machine learning natural language processing, it's not just one model or it's not just one algorithm, it's a combination of all the algorithm that should be in place to augment and help the business units to make decision. So part of that you're right. Sentiment analysis is part of it, top end modeling as part of it, temporal analysis is a part of it, even geographical analysis. That goes back to unstructured data analysis and couple with any of the structured data, you can imagine it's demographics. So for people can make that leap of association to be able to say, making this, Hey, I'm complaining about your burning smell in the vehicle. Now I know what are the demographics of vehicle buyers that it was in our CRM system that actually really reach out to, so we're marrying all those things together.

Jon Prial: Yeah, inaudible you're setting my next question. So, I'm an unstructured data guy. I spent many years in enterprise content management and all we really dealt with at the time was all this unstructured data and whatever associated Metadata we created. But now you're bringing together data from many sources that are classic unstructured data, but you have all of these structured sources. You just mentioned a number of them were the geolocation or timestamping or if somebody is in a hotel room, you know what the name of the hotel was. So as you begin to bring all this stuff together, how does that happen? How many channels and data sources are you beginning to deal with?

Derek Wang: Yeah, that's a great question. So one of our ability in our platform is right now already allow people to bring over a hundred different data sources into their own analysis. So they can kind of pick and choose what are the data sources relevant to them, and then start piecing all of those puzzles together. And one of the area that we're driving into is building out a data connector market sector so that we are even broadening the scope of where you can bring your dataset into not just limited to unstructured data, but any of your structured data as well. So the reason behind that is where we're start seeing, one of the key factor we're doing is ask human are good at understanding structure data. So our machine learning is actually helping people to structure the unstructured data so you can marry those things together.

Jon Prial: Interesting. So do you see this different now for each company, for each industry? So what are the learning that Stratifyd is going through as you build your more and more customers into your space?

Derek Wang: Yeah, that's a great question. And absolutely we have start seeing number one, industry- based differentiation points. So different verticals may have their own specific data sources or areas of focus they may have, which is naturally progressing to what kind of data they want to pull in. But also what we have started seeing a common practice across different verticals in terms of what they're trying to look at, right? Everybody's looking at customer retention and everybody is looking at predictive C- SAT, predictive customer satisfaction scores and etc. So where that helps us is to say, Hey, we can bring the best practices across different verticals and really helping those companies who leverage our platform to be the leader in their sector to enforcing a kind of customer driven benefits for the organization.

Jon Prial: Very cool. So I'll just got to end on a couple of future questions for you. I mean, for years we've seen analytic companies that are providing kind of simple, basic dashboards. Where do you see this going to allow more automation and kind of building some of this into business processes directly?

Derek Wang: That's a fantastic question. So did you speak into our heart of practice of our platform? So our platform is actually three layers, data, analytic and workflow, right? So my fundamental belief of AI is automation. So we're helping making the business units life more autonomous, and they're off shedding in a lot of their decision maker to help making process for AI to help on that. So definitely we are big in workflow integration. We're big in generating what I call a more AI workforce for each individual to automate their workflow and then helping them to reflect their decision- making in a more autonomous way and linking that into other workflow integration tools. Like we are in conversation with many other kinds of our partners, namely Microsoft or Mastercards that to bring the workflow automation generated by our platform to their client and their client basis.

Jon Prial: Very cool. So as you begin to bring third- party external data, because there's lots and lots of that that's out there, and you begin to bring it together with all the different data sources, number of those data sources might include companies private data, have you thought yet as a big SAS solution base to allow other companies to integrate some of their private data and look at new privacy, protecting techniques and differential privacy is a way to add more value to all your customers?

Derek Wang: Absolutely. And I think privacy and trust is number one thing as an enterprise software company that we abide to. So couple of the practice we're doing right, we are ISO 27001. We got many security certifications, happening in Stratifyd, but most importantly, we set up our infrastructure to physically separate the data for one company to the other. So there is no mix and match about companies data. We're very vigilant even about saying where the metadata go. So that's a very important thing for us. And part of the... You mentioned exactly we're working with Georgian partner on differential, privacy practice and amid that and team were building that into our product so that we can bring additional degree of confidence, privacy and security to our clients. And that continues to be a major area that I will personally look into and leading the charge of the product development. That's super important for us. And also I couldn't speak enough about the importance, not just the technology, security and privacy. It's also the human security and privacy, that's where I invest a lot of resource into Stratifyd. the trainings of the human level of security, the practices, how we hold against our employees for them to be always have the trust and privacy first in mind when we're talking to our clients, when we're presenting our company to them, everything is very important on the privacy side.

Jon Prial: Fantastic. Derek, a pleasure having you with us today on the podcast. Thank you so much. And look forward to seeing you again.

Derek Wang: Thank you most John, it's always my pleasure.