In Mountain View (CA), we meet founder and CEO of AgilOne, Omer Artun. He shares his story how he came up with the idea and founded this company, how the current business model works, what are the current trends in predictive and descriptive analytics market, as well as Omer provides some advice for young entrepreneurs.

The transcript of the interview is included below.


Martin: Hi, today we are in Mountain View, at the office of AgilOne. Omer, who are you and what do you do?

Omer: Thanks, Martin. My name is Omer Artun, I’m the CEO and founder of AgilOne.

Martin: What did you do before you started this company?

Omer: I used to be a marketing executive at BestBuy, and before that at MicroWarehouse, which was a billion and a half dollar computer and related products reseller.

Martin: Ok, so you have been an employee before and what made you switch to this new kind of career mode, from being an employee to becoming an entrepreneur?

Omer: My background, actually, all the way starts at a PhD machine learning, really fascinated with machines that can think algorithms and all that. And then I went into consulting side, I was with McKinsey for three years, doing strategy consulting. Specifically, I was really interested in applying science to marketing and then I finally found my calling, I said “I want to do this at a company”, so that’s when I did two stints at MicroWarehouse and then at BestBuy. But then, towards the end, I realized that there’s many more people that need to apply predictive analytics or marketing analytics because marketing has always been not a very scientific endeavor, and marketers are more creative people than mathematical people. But there’s definitely, as the digital world evolves, as you try to market to millions of people and you want to make that experience unique and personalize the experience, you need to be using math. You need to be using analytics to make sure that the experience is personalized. And I thought that the approach that I could bring, I could bring it to many more marketers than I was doing myself so that’s when I saw the opportunity in the market and started the company.

Martin: Ok, great.


Martin: Omer, how does the business model at AgilOne work right now?

Omer: AgilOne is software’s service, cloud based software company and we are a subscription model where our customers sign up for our service. It’s a yearly subscription that they pay us and they basically use our software through the cloud interface.

Martin: And do you have different type of pricing points or packages or…?

Omer: Yes, we do. So, we have both, from a functionality perspective, two different editions. There’s an enterprise edition and a digital edition, and those price points also change with the number of active customers that our customers have. Because, ultimately, like I mentioned before, we are trying to personalize the experience and have marketers really have a handle on understanding of their customers, as well as be able to kind of take action on a personalized level. Our whole focus of value creation is about creating more active, engaged customers. So, our pricing is based on the number of active customers they have. The more you have, the more our fee is.

Martin: And how did you come up with this pricing strategy or a pricing structure?

Omer: Pricing is an interesting topic. There are three ways you can do pricing.

  • First way is value based pricing. You look at how much value you create, and then you work backwards into how much of that value you can capture.
  • The second type of pricing is cost based pricing. So you can say “I have this much cost, I have to have this much margin, so that my operations cost this much and I have to have this much profit”, so you can basically come up with pricing that way.
  • And then the third type of pricing is competitive or alternative pricing. So, what are the alternatives, or what are the competitors charging, and where do I want to sit in terms of the functionality that I offer, and what the competitors’ alternatives offer.

So, depending on where you are in your company (building process), and where you are in the market (whether you are entering the brand new market or whether you are trying to enter commodity market), you have to take a lot of things into account, but you have to use either a triangulation between those three strategies or pick one of them. Ultimately, if you’re doing, for example, something super unique, that there’s no alternative for, you can basically do the value basic pricing and capture the huge amount. But if you’re doing, if you’re in the commodity market, selling grains, you have to do competitive pricing because commodity pricing dictates competitive pricings.

Martin: Ok, great. And can you tell us a little about the basic structure or the basic logic that is behind predictive analytics in the marketing segment?

Omer: Yes, sure. The predictive analytics take a look at the customer or whatever problem you’re trying to solve as a whole, and all of the interactions or elements that make up a customer behavior. So, it could be an online click that you made(but what you clicked), it could be a keyword you typed, it could be an email that you clicked with the specific header, it could be a product that you browse. So, you’re doing many things, and if you look at all of the things that we do in a day, whether it’s online or whatever, there’s a lot of noise in this data. There’s a lot of predictors in this noise and there’s a lot of signal within this whole dataset. And what predictive tries to do is, you need to both:

  • (i) learn from the past: For example, the way I would become intelligent is I would start asking you questions like how old are you, how much do you weight, do you run every day, what do you eat and all that kind of stuff. The answers to those questions are descriptive, I’m learning about you;
  • (ii) but once I ask those questions, now, as a human, I can basically take those answers: And I can also ask you what is the temperature outside, which is completely irrelevant, it has nothing to do with what I might predict about you. But knowing that you do x, y and z, I can predict “This person, you’re not very likely to go on expensive vine trips”, so you are more of a sportsman I expect you to go hiking. So I’m making predictions about you, which basically is using what I’ve learned from the past, and also eliminating the noise, which is the temperature outside has nothing to do with what kind of vacation you’re going to take.

So, the predictive technology, with all this data that we collect, it allows us to filter out the noise and basically focus on the signal, but also not do it in a way that looks backwards, but it looks forward. You are no longer just taking action; you can take action of what happened in the past, it’s already happened, you can only change the future. So, the predictive is really important in changing the outcomes that you have by knowing what is likely to happen.

Martin: What type of data sources do you typically use? Do you just use the data sources from the customer or do you enrich with other datasets as well?

Omer: We use the data both: from the customer, as well as from the outside. From the customer, we’re looking at the interaction data with their customers, so it could be web, email, call center interactions, the purchases, the app usage and those kind of things. And then the external stuff could be like where you live, whether you’ve recently moved, how much is your average household income, do you live in a high rise or a single family house, those type of things. A lot of those things also comes part of the decision and is very predictive.

Martin: And the algorithm, is it very unique for each and every customer and if yes, does it need some time for calibration of this algorithm or does it work ultimately from time one?

Omer: That’s the beauty of it. There are lots of tools out there that you can build predictive models, one at a time for different customers. But the innovation that we really made here is understanding the predictive process, there’s like an 8-9-step process that happens, but creating that predictive process in a self-learning environment, which means that we’ve set up this process and we expose the data to it and the system for every single customer will learn from the data and then will convert on to different predictions for different customers. In one case, for example one of the predictive algorithms we have is clustering; we group customers together that behave differently. If you’re selling vitamins, there’s a joint and heart medicine group of people, and if you’re selling cosmetics, there’s a facial cream people. There’s no way that I can dictate it, the math dictates it and from the data it’s going to be different groups of people for different kind of customers. And not only that, but also over time these groups kind of change and move, it’s a dynamic, it’s a learning system, just like humans. We learn and then we forget, and that learning plasticity is what we kind of built in our system, which is what’s unique, because now we can basically serve 150-200 customers, all of them have unique outcomes for each of their customers, but at the end of the day, we’re running one software platform.

Martin: And this is one of the differentiating factors, for example compared with other companies in predictives…

Omer: That’s true. You know, the alternative way of doing this is getting tools like SAS, SPSS, MathLab, those kind of things and getting bunch of data scientists and doing it by hand for each and every customer, which takes a lot of effort, lot of manual effort, lot of consultancy, and we basically do all of that in the cloud, in a self-learning way. It’s still using the same algorithm, but the algorithms basically get trained and fine tuned by themselves.

Martin: If I summarize your product from a customer perspective, is it correct to say your product will help me to minimize my marketing spending risk by knowing or expecting to know how much money I will spend for customer acquisition or for specific actions I want my customers to do?

Omer: It’s that, as well as being able to create a unique experience for the customer, so knowing where you are in the buying cycle, where you are in your customer life cycle. The way I go back to, I used to teach at NYU, for their MBA class, relationship marketing and the example that I used to give is all of the CRM and analytics, all that stuff that we’re talking about, it goes back to the corner of butcher shop experience that people had in the 1970s. If I’m a butcher and I have 50 customers, I know each and every one of them, what type of meats they like, how do they like different cuts, what’s the special days for them, how many people coming for dinner, all that kind of stuff, and now, equipped with that, I can basically customize the experience to every person, so you come in and I have bones prepared for you because I know that your dog likes them, that kind of stuff. When you take that experience and you basically say “Ok, now I’m going to take this” we went from that to buying meat in the styrofoam packaging in the supermarket, which was completely impersonal, so now the world became a product centric world, because you needed to mass produce, you need to grow and so on. But now with the advance of technology, which is free to customize, now I know different groups of people, now I know what they like or don’t like, now I know whether they are going to come to the store or not, I can predict these things, group people together. We’re trying to create that experience using data science, basically. Now, instead of handling 50 people, you can do similar messaging and interaction for millions of people. I know that, if you came and looked for patio heaters as a keyword on my website, and you looked at couple products and you didn’t buy anything, I know that you’re looking for patio heaters, you should get an email tomorrow, not about furniture, but about patio heaters, best sellers of patio heaters in your area and so forth, which might be different for different areas. Creating that response is what we enabled.


Martin: Let’s talk briefly about the corporate strategy. You talked a little about the differentiator. What is the driving force in creating more and more competitive advantage? Is it more like integrating, doing not only the predictive analytics, but also doing that kind of execution with CRM etc., or is it something else?

Omer: That’s true. At the end of the day, marketers care about results. And the results come from execution. Giving them intelligence and all of this sophistication around understanding their customers and being able to react with… Actually, rubber hits the road when you actually have that interaction with the customer at the moment of truth, when the customer is in your store, or the call center or opens up their email. You basically, the intelligence that you’re providing, the predictive, or whatever, informs and enables that interaction. So, you need to be at that execution later, to enable it, to show value to the marketer.


Martin: Let’s talk about the market development. You talked a little bit about descriptive marketing analytics and the predictive. What is the inter-relationship between these both, is there some difference in terms of marketing development?

Omer: The descriptive analytics, you can think of it as a business intelligence, reporting. People called it analytics. It’s basically reporting on what happened in the past. You’ve acquired this many customers, you’ve lost this many customers. And this provides, I mean, it’s very important, because that’s what all of the predictive stuff is based on, but it’s also very reactionary. If I told you like you’ve lost this many customers, you’ve already lost them, so now you’re… But if I told you here are customers you’re about to lose; now you’re like “Ok, now I can do something with it”, right? It’s actually turning the equation. You talked about the marketing spend effect that is where do I spend my money? It’s much harder to reactivate a customer after I’ve lost them and they haven’t bought from me for two years, vs. a customer that I’m about to lose, they are still opening my emails, but they are kind of not buying, they’re showing you signals that they’re going to go away, so I can do something about it. That’s where the difference in predictive and descriptive analytics comes in. You need both, but predictive analytics kind of turns the equation to be more proactive as a marketer than reactive.

Martin: From my understanding, predictive marketing analytics or predictive analytics in general is quite new topic. Can you tell us a little bit when this kind of market development started and maybe some numbers in terms of market size or market growth?

Omer: Sure. So, the market I think has started long time ago, with, like I mentioned, people using in-house tools and data scientists and superstitions and so forth. It existed really well developed in the banking industry and credit card industry for many, many years. It is on the predictive side. On the descriptive analytic side, a lot of catalogue marketers in the past used RFM type techniques or they looked at analytics, to figure out who to send catalogues to. So, all of that stuff existed, but now, the difference is, you now have customers that interact with many more channels with you. So, I can now go look at stuff, pin stuff on Pinterest, I can look at add in Facebook, I can look at a video on YouTube, I can type a keyword on Google, there are so many ways that we interact and that changed, so the amount of touch points have changed. The amount of data that’s being collected and available has significantly changed. So, before most businesses only collected like, who bought from me and who’s on my mailing list, and so forth, but now it exploded to I know exactly what keyword you typed in, and I know what product you looked, and I know what you returned, I know whether you called my call center and whether you were happy or not happy. All of that information is available. And, again, you’re trying to utilize that, to change the interaction, so you need to be intelligent. When I spend 50.000 dollars with you last year and I call to return a 50 dollar product, don’t treat me like the guy that returned the last three products they bought and so forth. People are looking for that, by giving up their privacy and this much data, they’re looking for something in return, something in return is the explosion of information that’s coming into the consumer, they’re looking for that information to be a bit more organized, a bit more personal and so forth. So, that’s why people are saying “Ok, we cannot handle this with rule based systems, we cannot handle this with…”, in the past, most of the marketing automation systems was like if the customer did this, than do this, if this, do this, that. So, these are rule based systems that don’t take into account the complexity and also don’t eliminate the noise from the signal, so then you’re really focusing on understanding that individual user. So, from market development perspective, this is pretty new, I would say people started talking about it, I mean, we’ve started a company 8 years ago and a lot of people told me like “What is this about? I don’t understand what you’re doing”. I knew it generated results, I’ve tested it at the big companies and so forth, so I knew that it was going to become mainstream, but really, it started taking off as a word, as a people’s interest and speaking over the last three years or so. And more and more people are adopting it, trying it, but in terms of the adoption cycle, if you look at people who are early adopters vs early majority, we’re still in the initial phases of market development, where people are, a lot of people are interested, a few more sophisticated people are trying it, and the people trying it are getting incredible results. We’re seeing results where, in terms of, for example, revenue per email, we saw one customer that was, that went from 3 cents an email, revenue wise, to over 13 cents an email. So, the results are incredible.


Martin: Omer, you collected or learned so much over the last 8 years. Can you share some of the lessons with our readers?

Omer: Sure. Everybody ask me what it feels to be an entrepreneur and what you need in an entrepreneur to be successful.

  • First of all, in my opinion, you have to be, these are my opinions, you have to be very rebellious. So, whenever you’re starting something new, that either doesn’t exist or you’re changing the way it’s done currently, you’re going against the grain. You’re basically trying to convince someone that this doesn’t exist, but this way of doing it is good. Or, you’re basically saying like, that way of doing it is, that’s how dinosaurs do it, and this is a real way of doing it. So, you basically have the rebellious attitude that you’re going to change the world. You really are out there, you don’t… When I started, people said “Well, IBM does this and these other companies do it, why would anybody go with two person company and not go with IBM?” So, you have to have the attitude and you have to be like, very tenacious is the second.
  • So the first thing is being rebelious and second one is tenacious, which is, you have that attitude where you just bite and don’t let go. You go after a customer and you service them, you make sure that you create the value and then they become your evangelist and you go get your next customer and so forth. So, it’s a knife fight and you’re basically fighting big companies and you don’t have the brand or the resources or anything behind you, so being tenacious is really important.
  • And the third one is being an adaptive problem solver, which is really important. So, all the time you deal with having lack of resources is, I think, the mother of innovation. Not having resources allows you to figure out, cut down to the most important thing you need to execute. In the initial phase, it’s understanding like, right now we can execute like 300 different campaigns. In the beginning I could only do 1, but which one is that is really important. And then figuring out the sequencing of a business, at every stage is different, like initially, I would do the selling, delivery, billing, and password resets, and then, after a while, I got to have someone doing the billing, so at every stage it’s really important to solve, being a very creative problem solver.

Martin: Ok, great. Thank you very much, Omer, for your time and have a nice day.

Omer: Thank you.

Comments are closed.