 Live from San Francisco, it's theCUBE, covering Informatica World 2016. Brought to you by Informatica. Now, here are your hosts, John Furrier and Peter Burris. Okay, welcome back everyone. We are here live in San Francisco for Informatica World 2016. This is SiliconANGLE Media's theCUBE. Our flagship program, we go out to the events and extract the signal from the noise. I'm John Furrier with my co-host, Peter Burris. Our next two guests are from Jewelry Television, George Willow, chief technology architect and lead nor director of business intelligence and enterprise data infrastructure. Welcome to theCUBE. Thank you. So, live streaming, you guys do live streaming, you also do cable, jewelry shows. So, talk about your environment. So, you guys have your own software, you have your own data. Share with us what you guys do from a technology perspective. From a technology perspective, we're always looking to build our in-house applications to really meet the tasks of what we want to do internally and also provide new services to our customers. So, for us, really the best way to tune that is to understand that data. In the past, we used to go through with a lot of statistical models, graph kind of style databases and build out the information that would guide the feature development in our software and we want to take that even further to the next steps. So, right now we're looking at ways that we can start incorporating things that really lend itself well to a live television broadcast for things like real-time feedback that we can even provide to our show hosts. So, we're very interested this trip into Informatica world to talk about live streaming kind of solutions and where Informatica is taking the product lines to support that. So, talk about the business. So, you have the television, so, jewelry television. So, you're showcasing jewelry. I should be a customer, get Linda some more jewelry. Linda, if you're watching, I got some jewelry coming your way, hopefully get a discount. But you have a product, a media product. But that's not just the only thing, you have that product, you have the business. Describe the landscape of what you guys are operating. And then how the systems and the software and the data fits in. Absolutely, we're operating the 24-7 television show. It's 365 days a year. We also do the live streaming of that. We have a great website presence and a mobile present as well. We consider ourselves Omnichannel and we like to have a consistent kind of opportunity for our customers to interact with us. With that comes different channels of data. Comes different application integrations. And we're very much a enterprise Java shop on the way that we design and build our applications. And we use the scrum methodologies on how we approach that. We also like to incorporate the data that we see into the designs of the applications. Things as mundane as default options that you're selecting within the applications themselves to optimize that experience. Because when you have a call center representative on the phone with a customer, you want them to be as efficient as possible in looking up information, finding alternate products or even dealing with availability of products for us. Some of the challenges we have being kind of on television with the live stream, we have limited sets of products sometimes and limited availability. Because each piece of jewelry is unique. I mean, we mainly have 10 of something. We mainly have five of something. So we need those real-time models to understand how many do we have left to sell or even have available. Because we don't want to discount or just, we don't want to really discourage our customers by not having product. Well, you're promoting something if it's not available. Yeah. You want to have that kind of agile ability to react and say, maybe we're sold out of that, but here's something very similar to that. Or here's another offer that might be right for you. We don't want to oversell that in any case. We don't want to disappoint customers. So you have the broadcast, you have the e-commerce, you have probably in-house fulfillment, all that good stuff, all kind of across the board. Which one is the under the most pressure for the systems and the data? You know, I think really the order management system is very key and that we're offering the right kinds of things for payment options, even in-house warranty options. But our warehouse fulfillment is been under a lot of pressure as we've been growing as a company. We're currently embarking on additional automation with a lot of additional hardware. We're also really automating those processes as well. And out of a lot of the tuning that we've done for internal systems, the data analysis, and even the patterns on how things behave have really helped us respond better and get product to customers faster. Even how we deal with our reverse logistics on how we have customer returns come back, we wanted to optimize that as well to understand those patterns. And what we see is commonalities when we start looking at our data between the way that our customers like to purchase from us and then the patterns even on the way the things may come back as returns. And it's a little different with luxury goods because you typically have kind of an industry standard, a little bit higher return rate. But the reasons for those returns can actually vary quite a bit more than if you were selling like electronics or something else. So a good example is a customer may purchase a size six ring and a size seven ring and choose which one fits best for them and return the other. So there's nothing wrong with that ring coming back in but we want to know that that's a common scenario coming back in so we can make a routing decision to send that to a cleaning service within our in-house jewelers. In other cases, we may have a collector kind of persona where they're interested in getting a really nice gemstone and they'll perhaps purchase three of the same exact thing, choose the one that best fits their collection or that best fits something where they're trying to build their own ensemble from it and return the other two. Knowing those kinds of patterns allows to route the way we handle that reverse logistics kind of processing in a more efficient way. And potentially turn that very action into something of a service that might offer some or be associated with some differential pricing. So you're an architect. Yes, that's correct. And you're over in the analytics side. So how do you see yourself then plugging into a lot of the business decisions that are being made to try to take what is currently an operational act and try to start turning it into a business service? Obviously, data and knowledge is power. So one of the things that we try to do is depending on where somebody needs to do their job but they're a returns agent and they're processing one package at a time, can we give them data to enhance their job when they're doing it? There may be things like George referred to for the pattern. Does it need to go to the cleaner? Does it need to go to the repair shop? If we can tell them that while they're doing their job, as they're doing their job, then we can do it more quickly and more efficiently and cheaper. You guys write your own software. So in terms of that, can you explain, take a quick second to describe why you guys have chose that route versus using other software? Absolutely, and in a lot of ways we have to. It's like there's really no concept of a good show planning kind of application out there that would understand how to schedule show hosts and products and promotions and things around it. So it's something that, it's a very core competency we have is to build that. Along with that is the integration that we get from knowing how that the show planning is working and which items are coming up and which are key items and be able to even integrate that into our warehouse, I'm sorry, our order management system on how we fulfill those. Good examples of that are knowing what item is currently on television so that it's a quick way for one of our call center representatives to quickly put that into the cart. Now the reason we want that so quick is we may have a very limited quantity of even a specific size and we want to be able to lock that in. We even use a reservation model to make sure that they are actually locked in and get that. That's a difficult feature to find and anything that you would buy off the shelf. And even within that model we offer the ability to even wait and try to lock in and get one as soon as it becomes available because it's really a short quantity available to get that. And beyond that we'll take it to even a wait list to where based on our projections of the data that we produce from our data warehouse we can know what an estimated return rate may be for this product and actually understand even better how to manipulate the way that we sell that or even offer that product to get the best outcomes with our customers. How about open source? You guys use open source at all? Open source is huge for us. So is freeware. Very much the Java platform is very key for us. Within the open source frameworks we're very fond of Spring Boot, Apache Camel, Apache Kafka is something that's starting to come in force as well. Everyone loves Kafka. That's what Jeff Frick said on Twitter. Absolutely. Kafka is everywhere, whole real time thing. It's pretty hot, right? And for us it's really that next generation of what we had is JMSQs. It allows us to really step up and scale that messaging fabric for us to be able to do things where we hadn't thought possible before. How about social channels? You guys seeing on the live broadcast and the affinity towards certain channels, Twitter, Instagram. Absolutely, Facebook, Twitter. We see a great deal of feedback from our customers and part of where we'd like to go is to be able to incorporate even more of that into the dashboards that our show hosts have visibility to. So that's one of those cases where we would really want the real time sentiment analysis of those streams of data, of unstructured data. Be able to say, are customers engaged? Are they happy with what we're showing? Are we hitting the right market segment? And even within that to understand what kind of group are we reaching? Are we reaching the fashion enthusiast? Are we reaching the business professional who's looking for the right kind of jewelry to complement our outfit? Are we reaching the collector? Are we reaching somebody who's wanting to buy maybe even a lower price point who's kind of younger, wanting to share that with their friends? So being able to understand that segmentation as well as the sentiment and the classifications of those is a very valuable thing for us. How about like listening engines? You guys have your own software there and you use off the shelf or listening. We actually use off the shelf for the listening engine itself that we have integrated for our call center. And right now we haven't fully tapped the potential of what we can pull in on that yet. On the architectural side, I'm sure you get bombarded. We just saw about the tsunami of buzzwords that's been polluting the industry. Hybrid cloud, public cloud, on premise. I mean, I'll see you guys are on premise, but thoughts about cloud, and how does that fit in? Absolutely. I think cloud has to be there always, you know, as we look at any solution that we're going to implement. And for us really, the cloud is the agility of deployment. And that was a big deal for us. Part of what we're here about was we are actually invited to attend the award ceremony as one of the spotlight nominees because we're really looking at how do we really think about optimizing the way that we do our internal things and how do we prepare ourselves for potential cloud plays that we have? We're already in the cloud in some aspects. We have our website is effectively cloud hosted as well as our streaming service. It makes a lot of sense for us to provide that stream from Knoxville and then using streaming services to distribute that. And then we also have our caching layers like Akamai out there as well. You know, and that's very present in the cloud for us. We still have a lot of cloud to ground kind of things going on. And with that, we're looking to even optimize those in some of the plays that we're going to be doing in the near future as well. I wouldn't be surprised to if you see even more of our infrastructure systems moving toward cloud. But it's for us, it's that provisioning aspect is so key for us in the management aspect. We've been able to achieve a lot of that within our own data center using a lot of the VMware product lines, a lot of the automation tools where we actually build and produce things. Part of what we did with the Informatica products was actually to take the product itself, run it through an actual internal packaging process to produce RPMs where we could do literally a distribution across using a local RUM, a young repository and do literally one command installs of a full Informatica environment. There's a ton of savings for us and being able to bring those environments up quickly. And that was part of what we looked at with Informatica because we'd gone through some growth with Informatica for like the last three years. And within that time, what happens is we have different consultants, we have internal parties working on this and things can happen when you're dealing with on-premise kinds of things where tweaks can be made and the reason behind the tweaks lost over time. So what we wanted is we went to our 961 upgrade was to go in, understand those and even go back and research some of these tweaks, incorporate that back into kind of a roll forward strategy so that we not only have lineage of our data but lineage of our distribution artifacts and we know who changed what, where, when and we have it stored even with our software repository built with our Jenkins internal continuous integration engine and delivered right into our young repositories where it's one command to install an Informatica power center, even power exchange and environment. And that was a big deal for us. Yeah, well, you're not going to be able to do this competitively on-premise unless you can do some of those things. Exactly, yes. So only coming to some of the things that you're trying to do. Clearly, jewelry TV is kind of a combination three or four different businesses. There's the TV business, there's the e-commerce business, there's the logistics business. How are you working with all these different businesses to prioritize their needs globally? Number one, and number two, what role is Big Data playing as you try to prioritize the priorities? One of the things we've been very lucky to have is we have very great buy-in from our business, especially business leadership. So we have a steering community that helps us with a lot of those prioritizations. So as they come in, we get feedback from all the different areas and we sit down collectively to decide which ones are we going to go after and which ones we're going to go do. Specifically to the big, big data question, there's so much data that we have at jewelry television. A lot of stuff tends to be around interactions with customers, but there's some interactions with customers that don't necessarily result in sales. So while the company does a lot of things, the one thing we'd add is we're also an entertainment company. People watch for the fun of watching. We're driven by sales, that's our big, big deal. And those other pieces of data where we offer things to customers or we have things that will engage a customer but don't result in a sale, what do those mean? And what are those actions? And how do we get less of those actions that don't result in sales? But again, it goes to that steering community process and we partner with our business users and we really sit there and we help, we bring our input, they bring their input. It's not just technology drives it, it's not just they say we've got to do this. It's really a collective partnership on how we decide what we're going to move forward on in terms of data and what we are going to put in front of them and how we're going to put in front of them. So the traditional media model is, if you take HBO and some of those guys out, is to create content and programming that is supported by advertising. You guys are creating content that's supported by actual transactions. That's correct. Are you also being somewhat supported by advertising at the same time? I'm not familiar with that. We are not, we are different than your traditional content provider in that we pay for air time. So the people don't pay us to broadcast our content. Got it, so that you guys are generating content, paying for air time to get your content out and using the transaction to monetize. How is that, therefore, how is that as a business different from a media business that is monetized mainly through advertising? And what role does big data play in making you more effective as opposed to an industry that's really trying to use big data to tell the advertiser something? That's true. We obviously don't have to worry about ratings or anything like that. But the things we look at are, we get 100 orders in a minute but we had 10,000 people watching. That may not be as good as if we had 1,000 people watching. So we look at that and that's what I was talking about. There's opportunities and we try to see what happened where they did not engage or that we started an engagement but an engagement did not end in a transaction. And we will look at things like host behaviors. We'll look at things like product sequences. We'll look at stuff like that and say what is a winning recipe? What is a winning combination for us to help this first engage with the customer and continue that engagement all the way through a sale and then again to subsequent sales and hopefully many subsequent sales. That's, you use the term host behavior. That suggests that you guys are also at the vanguard of actually thinking through some of the kind of the human side of the interactions. How someone presents themselves on the screen, generating conventions that actually can be measured. Have we got that right? Yeah, absolutely, absolutely. To tell us a little bit about that. Because I mean, every business is thinking more about how do I, what conventions do I use to say something about my customers that I can then turn into data? How are you guys doing or thinking about it? Well, one thing we know specifically to the host example is we know viewership. And then obviously we know the sales. So we can correlate viewership to sales. Sequencing a product goes into that. What, how hosts perform? Are they sitting down? Are they standing up? Are they smiling? Are they by themselves? You're capturing all that information. Yeah, we know all about that and we're able to build essentially winning recipes. We actually call them plays, if you will. We go out there and say, if you do these 10 things in succession, you'll be more successful. I mean, if you did those 10 things individually or if you did number three as number seven versus number three. So we do, we do know. So what should John do differently? Don't ask any questions. So what are the questions, take one. Keep them, wear glasses. We saw the glasses increase the ratings. Makes me look smarter. I mean, that's a good point. I mean, this is something that we live every day because the data and the player, sometimes we're free, so it's all free. So we got to look at, we don't have sales driving in the queue, but this is an example of mobile apps. People think about A.B. testing. It's all software. Yes. And so we think about this all the time. We have our own listening engine we built and we have our own crowd chat application which is our own whole chat using hashtags. But we're constantly looking for advice. What advice would you give us? Boy, the best thing is to get it as real time as possible and as accurate as possible and measured as possible. Those are the real challenges when you're up against the live television broadcast. And the more that you can do that with these newer streaming technologies, the more successful you're going to be. And the more that you can augment that live stream information with what you've already done kind of in the back house with your big data, the better off you're going to be. Do you guys stream through a service so you have your own CDN? We actually use a combination of Akamai and also Amazon Web Service streaming. I want to come back to what you said because you said something very important and I want to lose sight of. We here at SiliconANGLE TV, we're going out and generating digital assets that companies can then use and consumers can then use to drive their engagement models better. And you are also trying to find ways to take something that's analog, like a host doing something, but translate it into data. You're trying to digitize it. You are in many respects at the vanguard of trying to digitize behavior, both on the consumer side, but also on the host side. Translate that host into content that then can be analyzed. That's right. And then you don't forget e-commerce. And with obviously, yeah, very much the e-commerce side. There are, you know, that is a fascinating, you are trying to digitize things that aren't normally thought about for digitization. Do you have a lab that does that? Do you, I mean, obviously as an architect, is the architect's role to also start thinking about how we represent people digitally? Somewhat. And we actually have internal groups for a digital media group that focuses on that very specific thing. And that's a lot of where of our business analysts kind of statisticians are there that help us understand even what they're measuring within the data, even relative to things like camera angles, like every single one of our cameras actually has a live stream of information about, we know which one was turned on, when, how it's behaving, and what the overall experience is for a customer and how they respond to that. And you see that it varies more than you might think. It's IoT as well. It is very much. It's got everything rolled into one. The cameras are all instrumented. Oh, we have a little bit of everything. We have the cameras instrumented. Even at our warehouse, we can watch the conveyors. We're actually putting in more conveyors. One time we even were noticing slowdowns on the conveyors within the service times on the way that things interacted between barcode scans. And we could actually predict that we had a failure on one of the rollers. You know, that's pretty interesting. It's exciting. Very exciting. Thanks for coming on theCUBE. Really appreciate it. I'll give you guys the final word. What's your thoughts on the show? What excites you here at Informatica 2016? Everything, but more than that, it's meeting with our partners. It's meeting with Informatica and our representatives and really seeing these opportunities because we've been kind of branching into these worlds of the open source and saying, this is great. This Kafka stuff's awesome. You know, this Flink stuff is awesome. We have these proof of concepts that seem to be working. How do we bring it in? How do we manage it? How do we make this work for the long term? And it looks like Informatica is the right partner for us. And you guys were spotlight award winner for the innovation award. What was that about? Yes, that's correct. And part of what we did there was really the optimization in how we took the Informatica products and packaged it in a more efficient way for us to deploy it quickly and kind of bring our infrastructure up to par with more something you would expect on kind of a cloud deployment. But even beyond that, we were up against some challenges because we do the change data capture with the Power Exchange product. And we didn't have a good way to measure, are we getting the right throughput out of this? Do we have bottlenecks in our network? So we actually authored an open source tool called Asim throughput test that allowed us to go in and actually measure across our network and really optimize those flow streams that we were getting better than even published results on those kinds of things because the importance of us that those live streams of transactions and things are so high that we literally want to go through and pre-optimize that as best we could for our infrastructure. Any consistency and stability and speed, honestly. Absolutely. So we have across environments where we have our QA environment where we're testing and making sure things are working. And what you're seeing, orders of magnitude differences in there, you have to go and investigate, why did this happen? What can we do better? How can we tune this thing? And doing that, we're able to tune multiple layers all the way from the database side on producing the various logs that we're consuming through the CDC. We're able to tune the network. We're able to even tune the operating system at a TCP level to actually increase throughput as well. So you guys just aren't sitting back waiting for vendors to come in and give you software. You're building your own tooling to meet your needs using open source, but yet taking the goodness of, say, commercial software. Absolutely. And then even giving that back to Informatica. The things that we do, we share. We treat Informatica like a partner and they treat us very well as well. And I think that's the best ecosystem. The good business model. Absolutely. Thanks so much for coming on. Jewelry TV here on theCUBE. Thanks so much for sharing the data. Get building your own tools. That's the theme, of course. We do that too. So we know that's the way to do it, but don't get too over your skis, as they say. Absolutely. It's theCUBE live in San Francisco. I'm John Furrier with Peter Barris. Thanks for watching. We'll be right back with more after this short break. Hi, this is Chris Devaney from DataRobot.