 Live from Boston, Massachusetts, it's theCUBE at the HP Vertica Big Data Conference 2014. Brought to you by HP with your hosts, John Furrier and Dave Vellante. Hello, welcome back. We're here live in Boston, Massachusetts. This is theCUBE, our flagship program. We go out to the events, instruct the students from the noise. I'm John Furrier, the founder of SiliconANGLE. I'm with Dave Vellante, co-founder of Wikibon.org. And our next guest is Ernam Inham, head of marketing and business intelligence at Peak Games. And we're here at the HP Big Data Conference. Gaming is the hot market. I'll see, for us big gamer fans, we saw Twitch get sold for Google for a billion dollars, which was like a big thing because we know how great that product is, knowing the founders and everything. And just gaming in general is great. And big data is a big part of it. So, Peak, welcome to theCUBE. Thank you. Peak Games. So, tell us about what's going on because we always love to talk about games because Dave and I, Danny Ryan, co-founder, crowd chat, was at Riot Games back in the early days and built the back end for the Hadoop stuff. And so, we know the massive amounts of opportunity to do stuff with data, but also the massive data you're getting. True. So, talk about Peak Games, the environment. Just lay the land a little bit. What's going on there with the game, the infrastructure, and what you guys are doing. Sure. So, Peak Games is a four-year-old company. We're basically building up games. We started as a regional player in Istanbul in Turkey and then quickly expanded into other regions. Middle East first, then we decided to go even further to Western countries, to United States, to North America. We now have more than six million daily active users, which is a huge amount of data, as you can imagine. And this data is flowing us every day. And as you can imagine, the free-to-play business model is usually depending on a very small percentage of these daily active users being paid users. So, our revenues are coming from a very small percentage of these daily active users being engaged players, then paying users, becoming paying users. So, we need this data. Like, among millions of apps in the App Store, if you want to make a difference, you need to iterate your game so fast and so in a very dense way so that people can get exactly what they want from your game. Exactly good amount of challenge, good amount of joy, and then they become engaged users and then they become paid users. So, it's a classic freemium model. Yeah, exactly. People access the game for free and the people who get booked want the additional features. They have a community, they want certain things. True. And they're already bought in. That's kind of your model. Exactly. So, knowing when to convert people and offer them, is that an issue? It's usually like, if you try hard to convert people to pay, it's usually backfires. Because people can easily understand if you're trying to get money from them, right? So, our primary aim usually is to get them engaged, to make them enough fun and challenge at the same time so that they become an addicted player and in the end, eventually, they might end up paying. You know, we were talking about this earlier with Peter from Yummer, big data science guy, data nerd. Growth hacking. But you're basically referring to as growth hacks gone bad. Users now are smart, they can smell a growth hack. Import all your contacts. So, people are now fearful. What is the best way to handle that in your opinion, based on your experience? Because loyalty matters in the web but also people can get pissed off just as fast as they can get loyal. So, what's your experience in growth hacking and how to do it right? Well, it's getting even harder every day because we're not in the days where, you know, everyone posts on their open graph, like on their feeds on Facebook about the farms that they've built on or about little things that they play on in the games. So, people are a lot more cautious in what they're posting on Facebook. So, what we do usually is try to integrate all the social interactions within the games in a very contextual way so that, you know, when people post something, when people invite something, they get a real value out of it. And that's pretty much what other colleagues in our industry are doing as well. So, if he wants to have something from the user, we need to give something to it because no one would just invite their friends because they like the game because you need to give them something in order to get, you know, a value out of them in terms of their social interaction. What do you think about Zynga? Zynga also had a big run when public CEO got ousted, new leadership came in, they're free to play games. Did they do it right? Did they ever play their hand? No pun intended, even if they had. Texas Holden was one of their best games. What do you, what's your take on Zynga? I mean, still people, still huge numbers, but I mean, what's your take on that? Well, I think Zynga is one of the examples how fast our industry is changing. So, like leaders of three years ago are still very big players, but it is subject to change. And people get, you know, like in gaming specifically, you need to really iterate, you need to really innovate and evolve your games in the ways people like to see it, you know? And that's how I understand it from Zynga. Texas Holden poker is probably one of the most stable games that people can ever see. Like, you know, there's a great amount of people that are still playing it for almost like four or five years now, but, you know, the new games, the new tries of Zynga, I think, you know, it is showing us that there's still a lot of opportunity for new players in the industry so that, you know, everyone can build up great games and, you know, replace all the incumbents in the industry. So you talked about the 80-20 rule, it's probably more like the 99-1 rule in your community. And if I understood it correctly, you're using data to try to better understand that small base that's paying. Are you also trying to understand how to convert maybe some of those other non-paying customers, or is that just never gonna happen? I wonder if you could talk about that a little bit. Yeah, there's a small group of people who's really willing to pay and who already probably pays when you're looking at them, right? But there's another circle, a secondary circle, who is likely to pay because they're super engaged and if you create those moments for them to pay, they probably are gonna pay. But there is another outer circle that is never gonna pay, but you still need to keep them in the game because they trigger the inner circles to pay. So a paying user never pays because he just wants to pay. A paying user pays because he wants to defeat that non-paying user because he has been defeated by him for like a couple of times before. And that's how you trigger the paying user behavior. So the non-paying users are just a column fodder for the paying users to destroy. Yeah, I mean, there are some cases that you really need them to trigger in-game behavior. They're pawns. Kind of, yeah. It's a kind of game theory. How have you used data to improve the user experience or do you use data to improve the user experience? So for us, we need to personalize the user experience inside the game because whenever, we have a product, right? The product has three different features. We have a strategy game that you build up your city and attack to your other friends in the game. So this game has a couple of features. One being a city-building game, one being a battle game, and one being just a farm game. So we need to understand which user prefers which type of this game and alternate the tutorial, the in-game content accordingly. So if a user doesn't want to battle with anyone in our strategy game, we probably won't show our tutorials, battle parts in the game, but just show them a decorative item, the city-building part so that they can just go in that way. The same thing for the battler type of people. How much data are we talking about here? Well, currently we're holding 13 terabytes of data, but it was like nine, two months ago. So it's becoming huge, it's becoming bigger and bigger, and we're probably one of the conservative ones in the industry. So you can hear a lot, lot more data. You're in marketing, right? Yeah. At least part of your job and your current role. And there's a big discussion in the industry about how marketing is driving all the spend in technology. So I wonder if you could talk about that within peak game. So are you the guy with the budget? Yeah. How is that shifting? How are you spending that money, that technology budget? Well, we had the privilege to have the user acquisition at a very low cost a couple of years ago. So we had the privilege where no one else was advertising in our countries as gaming clients, but now everything got super difficult. So we need to look at every little cent that we're spending because the prices have increased due to increased competition. In Facebook and mobile ad networks in everywhere, simply. So that we need to take care about every penny that we're spending. And this, you need to be able to understand who's the big player in a particular country, who we can use alternatively as a B plan, as a C plan, as a B plan when we're out of the resources in Facebook and Google or in any other ad networks. What kind of successes have you had with your analytics projects and how do you think they compare to other IT projects? I mean, it used to be the line was that, I don't know, some huge percentage. I think at one point somebody made it up on gut feel, but it was like 40 to 70%, I don't know what the number was, of IT projects fail. You don't hear that same type of complaint with analytics projects. And I wonder is it because the business is more tied into it, but what kind of successes have you had? What kind of success rate do you have with your analytics projects and your big data projects? Well, it differs because currently in gaming, most of the analytics projects born out of a need, an actual need, because the game starts to fail at some features. The game needs some analytics input of that particular point. So for example, about one of our latest project is about matching algorithm of the game. So we needed to find a sweet spot in the win rates of the users so that people get enough joy versus enough challenge. So if you lose a lot, you turn a lot, right? It's like a slot machine. Exactly, so if you lose a lot, you really get frustrated because you just want to win a bit more to keep playing the game, right? And if you win a lot, you get, okay, like this is boring, right? I want to challenge myself. So we worked on an algorithm that improved the gap, that narrowed down the gap of the win rates among the whole distribution of the users. So it allowed us to make everyone pretty much on the same level in terms of the win rates and the lose rates. And, you know, believe it or not, like the next very next day, the number of games played by each user doubled. And like, it is a good part of gaming, I think. Like, if you're dealing with hundreds of thousands of people, every little change you make, every little detail that you're looking into can result in a, I don't know, double-dravening, tripled engagements, and stuff like that. Huge numbers. Exactly. So that's a question I've got to ask relative to the numbers is, value to players versus risk of disruptive operations? Because let's just say Hadoop for instance, say you want to throw a big Hadoop cluster around there. If that's not in your production environment, you're at risk of managing the relationship between what you got in production and what's actually good for the big data. So of course you want to store every interaction because you might need that data, a little, you know, certain headshot if it's a first person shooter game or if it's a certain move, you got to know everything, you got everything, so why not store it, right? So how do you deal with that? How do you deal with the need to store to create added value, either user experience, play experience, or revenue, or all the above, versus disruptive to the operation, meaning crashing? That is true. I think that's one of the most, you know, arguably topics in our industry. So what we usually do is like the data that you always need is probably very recent data versus all the historical logs you probably don't need that much while running the game. So I think the frequently used data that we keep, you know, very tight that we use to run the game and at that specific moment, like a matching algorithm, like the content that the user is getting, we use it very recent and very tight and then, you know, storing the rest of the data probably less frequently used places, or we can just add all, you know, go in and check in the data. So that's still a problem though, like for many gaming companies. So I've got to ask you, the gaming industry in general, where do you see it going in terms of the, you know, free play games, call of duties of the world, the consoles, I'll see mobiles of changing the game for everybody, right? So how is it all going to come together? All integrated in, you see the conversion, you see some consolidation, what's your view? I think cross-platform would be probably one of the biggest things that's happened. We already started seeing it, like World of Tanks, like it goes in mobile without any problems. We'll see a lot more cross-platform games. We'll see a lot more cross-platform experience and I believe like we started learning, quantifying the fun that people are having. So that's the good thing for us. As gaming people, we didn't really look at the data before, like 10 years ago, no one was just, you know, interested in the data that's coming from the players, right? But now we track almost everything. It's not working that much at the moment, but it will probably work a lot better in five years. Yeah, I mean, certainly horsepower for speed, the vertical thing will help, but also DevOps, right? Having a cloud, spinning up resources when you need it, versus having the over-provisioned bare metal. Are you guys all provisioned in-house right now Is there cloud? What's the infrastructure look like? We're like, most of the infrastructure is in-house because we have to develop it ourselves because the speed of the data growth is super high so that we need to have it in-house and it's a critical asset for the company as well. That's why we took it in-house. Okay, great stuff gaming here on theCUBE, extracting the value, extracting the signal from the noise. This is theCUBE, go to siliconangle.com, go to wikibond.org for free research. Of course, siliconangle.tv was a home of theCUBE. I'm John Furrier, Dave Vellante. Go to crowdchat.net slash hpbigdata2014 to follow the conversation. We'll be right back after this short break.