 Hey, welcome back everybody. Jeff Frick here with theCUBE. We are in our Palo Alto studios for a CUBE conversation as part of the Western Digital Data Makes Possible program. And it's very gracious of Western Digital to sponsors to go out and talk to a lot of different companies that are doing a lot of cool innovations. At the end of the day, it's all powered by data. At the end of the day, all software is just an algorithm sitting on data with a nice display for a specific solution. But this is, this one we're diving into sports and there's so much going on with sports and technology. And this is a great company that's actually been kind of flying under the radar for 10 years unless you're into the space. But we're happy to have him. It's GumGum and we're joined here by Brian Kim. He's the Senior Vice President of Product from GumGum. Brian, great to see you. Thanks for having me, Jeff. Absolutely. So for the folks that aren't familiar with GumGum, give them kind of the quick overview. Sure. So GumGum's an artificial intelligence company with expertise in computer vision. And what that means in kind of common language is that we focus on building algorithms that allows computers to identify what's happening in imagery. And then we apply that into different businesses that we feel like the usage of computer vision could essentially automate, scale, or drive significant value to those different businesses themselves. Right. You guys have been at this for a while. You're almost 10 years old. I think you said your 10 year anniversary is coming up. Yeah, we were founded in 2008 and a lot of the core teams still together from the original team that worked there. And we were doing computer vision before computer vision was probably sexy. Right, right. We've been working on it for a long time. We've built a lot of expertise around that. And with a lot of the improvements that have happened in machine learning and AI these days, it's just been kind of the big hot thing that has continued to accelerate and helped us grow significantly over the last few years. Yeah, all of our social media feeds are filled with the picture with the Chihuahuas and the blueberries, right? Trying to figure out which is which. But you guys have a very different approach than kind of what we read about now in the popular press. You built computer vision capabilities, but you built them for a very specific application, not just as a generic kind of computer vision. So I wonder if you could tell us a little bit about your strategy and how you guys got started in the ad space. Sure, so to your point, Jeff, I think a lot of companies have focused on what we like to call AI as a service. And what that means is that they build the capability of using computer vision, but it's really up to the end user to use it as a tool into their actual business and figure out where it'll actually apply. Our path forward has been focusing on building what we like to call full stack vertical solutions, meaning that we decide to go into a specific industry or division like ads, for example. We build an ad solution that's using computer vision itself and we actually sell that ourselves to agencies and brands direct. And then we work with the publishers on the other side of the coin to actually deliver those ad experiences and continue to kind of build our businesses around that model. Right. And how is a computer vision AI driven ad experience different than the alternatives? Well, I think there's a few things that we've focused on that make it different. The biggest, I would say is the placement of where the ads actually are themselves. So compared to your standard display ads that are around the content on a webpage, what we've focused on is actually building ad experiences that are within the content itself. What we call in image ads. And that's an ad that overlays on top of that photo editorial content that's on a publisher website. And what we've done there is we've applied our computer vision to understand what's going on in the actual images. And we leverage that tech to be able to then contextually match the ads from our clients to the actual pages themselves so that they're completely relevant to the page. And we make them look very slick in the kind of the design of it so that it feels very sticky and natural to the way that the page is actually designed and built. So what's different just so I'm clear is that unlike a typical kind of an ad placement in a page which is going into a dedicated spot as soon as the page loads it takes some data and it goes to the auction. You guys are actually looking at the page as I, the consumer of the page, I'm looking at the page. And based on the things that are loaded regardless of where I am on the page top middle bottom that's the stuff that drives your ad placement. Yeah, perfect example would be you might be reading an article about cats and PetSmart wants to advertise with us. So we'll understand that the image itself is about cats and put like a cat food PetSmart ad that's placed within the actual image itself. And then you might scroll down a little bit farther and find another article that's about a completely different topic. And we would actually match that to the relevant advertiser that fits that example as well. So how are advertisers measuring the delta in the value? I mean, it's some of the stuff in researching for this. I saw a great quote that you guys in your ad delivery want to be seen and frequent and respectful which I think is a really interesting way to take a point of view because clearly ads run a risk of being way too obtrusive popovers and pop-unders and pop-ups and I go to a page, I'm a huge customer within two seconds they're asking me if I want to buy some and I'm already your biggest customer. So how are the content publishers measuring this different type of an engagement with an ad presentation the way you guys do it? Yeah, I think what you talked about is the right approach of how we want to go to market which is that we want to provide a premium offering that's high impact but not obtrusive to the end customer and provides value to both sides of the ecosystem. So to the advertiser, it might be a premium CPM that they're paying in order for that ad placement but because of how relevant it is and how viewable it is because if you think about where your eyes are on the webpage, you're not looking at scanning like the sides of the page you're focusing on the content that's going down, the image is right in the middle of that so if an ad pops up right in the middle of that image it's much more viewable than say all the ads that are typically around the content itself. And on the publisher side to kind of the infrequency we don't want to blast an ad on every single image we want to match it to the most contextual and right placement and then therefore to the publisher when they get an ad that ad is actually paying out a pretty significant price point back to them and then they're happy with that experience because their users are not saying I'm seeing 50 ads on one page which is kind of the traditional problem that I'm dealing with right now. So that's ad tech market and you guys been doing that for a while but the reason we reached out to you specifically is your activity in sports which is a relatively new business for you guys. So how did you get into sports? How did you guys identify this opportunity and then we'll dig into it a little deeper? Yeah, so Gum Gum as a whole has always been looking at different ways that we think computer vision can be applied in a myriad of different industries and the opportunity that about 18 months ago we identified was that there was a very legacy business that was being built around providing media valuation around sports sponsorship. So what I mean by that is how do you quantify the value of a sponsor showing up, a state farm for example, showing up in a basketball game whether that's an LED placement, they're on the basket stanchion arm or they're part of the halftime show and the measurement that was being done traditionally was essentially done by people. So people were watching those clips, they were timing how long the different brands were showing up, they were measuring how big those actual placements were and they were then calculating some value off of it and we really thought that computer vision can one, automate that entire process. So take the humans out of the loop and get it to a point that it's completely automated and you don't have to have people involved, we can deliver it faster. So a computer can do something a thousand times faster than a human being, it can probably analyze it. And you're providing much more accuracy and efficiency of the actual data that you're providing back. You know the exact dimensions pixel by pixel that a computer is telling you versus a human being trying to eyeball where and when certain ads are showing up. So what we've done there is then built a business that is now called Gum Gum Sports where we provide media valuation to sports sponsorships. So both on the team side which we call rights holders and on the brand side. And we're essentially the middle man who's providing third party reporting to both sides so that they understand the value of what they're getting across their sports sponsorships both on digital which is essentially broadcast TV but also across social media which has been a huge gap that nobody's addressed today. So just before we go into the impact so before it was just a person they're watching the game and every time that state farm ad pops up on the stanchion or thing they're writing down approximately how big was it? Could I see it? Was it blurry? Was it moving? Was it in the center of the side? You guys obviously that's just right for algorithmic treatment. Because like you said the pixel so you're doing time, you're doing placement, you're doing quality. You've added a number of things beyond just simple that is there in terms of metrics to measure. Yeah so we look at things like where's the action happening? So we can identify in a basketball game where the actual ball is that's probably where people are focused on because the action's happening near that ball. So the closer you are to the ball the better score that you'll possibly get to your point how clear is it if you're panning back and forth through a game of your logo showing up? How big is it? How prominent is it? And we factor that into what we call our MVP factor or media value percentage which helps calculate what that end value looks like to the client. And was the demand driven on the supplier side or the buyer side? Were they looking for validation of this money? Was it value or was it, were you saying it was both? Yeah I started to interrupt you but I would say both sides were looking for somebody who's not favoring one or the other to give you validation. So they wanted an arbitrator who would basically say this is what I think the value is in the ecosystem so that both sides know how to negotiate when they want to put together their deals next year. The value to the teams are the more value you can generate obviously the higher that they can increase their price points. And I think to the brands what they're focused on is how do I optimize my ROI by the placements that are generating the most value for me and not waste my money in other placements that don't generate value for me. Any big surprises in terms of the value of a stanchion versus the value of a halftime show versus the electronic thing on the scoreboard? So I think the more than the placements themselves I think the biggest surprise that we found was how big social media actually has become in the valuation of all, of media in general. You know, there's a lot of talk about subscriber numbers going down on TV that broadcast is declining, that nobody's watching. But Super Bowl is down, I think this year, right? Yeah, Super Bowl is down this year. Which doesn't happen very often. You know, live sports is just not what it used to be. But the reality is I think consumers have just changed their habits of where they're consuming that content. So instead of having to sit in front of a TV for two hours they might go check the highlights on YouTube. They might go look at their Instagram stream and see a bunch of posts that are coming from, you know, fan accounts that they're following that give you the highlight clip. So being able to measure that piece of it that nobody's done before, what we found is that that value is actually as big, if not bigger than the broadcast site of it, which nobody has really quantified to the state. That's really interesting. So you're what sniffing hashtagged or something around a particular event to grab that data. How are you grabbing all the social data around, say, a basketball game? Yeah, so that's where the computer vision actually gets applied. So we don't even need to look at specific hashtags or specific accounts. We can look at the full stream, literally all of social media that's available publicly. And we're able to- Just plug into the API. Yeah, sift through all of that with computer vision and say, oh, this is not a sport. Oh, this one happens to be a sport. Now I know that it's NFL. Now I know it's tied to the Super Bowl. And then you can now classify all that data and then figure out the actual posts that you want to analyze and the ones you don't want to analyze. It's so interesting. I can't help but think back like to the Grateful Dead, right? Back in the day they were the only band that would allow people to record at the concerts, right? There was this huge, you can't record, you know, no pictures. And then they would trade the tapes, you know, in the parking lot before the game. And you saw that too with a lot of professional teams, you know, no phones. Is that the concert? Yeah. They're day and they're like, no phones. They're like, no phones. I mean, that is the way that people experience and expand and amplify these live events. And it sounds like what you guys are doing is really validating how important that is to all the people that are participating in that live event. I'll give you a perfect example with, you know, the NBA All-Star Weekend that just happened in Los Angeles recently. You know, if you look at the slam-dumb contest, half of the All-Stars that were in the crowd had their phones up and are basically recording, you know, something that they're probably going to post on their social media account later. With massive, massive, massive all. Yeah, each one of them might have, you know, two to 10 million people who follow them. So you multiply all of that. And that's probably a bigger audience than actually, you know, who tune into TNT or whichever channel that happened to be watching live, the actual slam-dumb contest itself. That's crazy. So I'm curious to know what the response is as you come back to this data. Obviously, it's great news for the publishers, right? A bunch of value that they didn't even know they were delivering. No one's even capturing. At the same time, I would imagine the advertisers are thrilled to actually see that they're getting this whole nother tranche of activation that they had no clue or at least no way to measure. Correct. So that's been the biggest surprise to everybody is how much value has been unlocked to them. And both sides are thrilled about it because now they can start to measure that on a consistent basis. And then moving forward, they can figure out how that fits into their overall plan for, you know, whether they want to charge more for their sponsorships or whether they want to price certain things in like social media that they never did before. Right, right. So my mind's going all kind of places. So can you, could you on sniffing that feed, find, say the State Farm logo, stay on the same thing? Where's State Farm logo showing up in a billboard that's on the one-to-one that happens to be in front of a pretty spot where people take bike paths? Are you seeing or even attempting to look for other kind of secondary social impacts of other forms of advertising outside of your core solution within the sports? Yeah, I mean, we've started to get feedback of people who are interested in solutions like that, whether it's digital out of home, different kind of businesses that have built themselves around wanting to track this type of ROI. And we've looked at a few use cases and talked to a couple clients that we're starting to dabble in now that might be interesting for us to build new businesses around just like the use case that you talked about with the digital out of home example. Right. Another one of my favorite lines that gets thrown around a lot these days, right, is in God we trust everybody else better bring data. So I'm just curious as to the feedback you're getting from both sides of that equation within the sports application, but now we have this data. I mean, how is that impacting people's evaluations? How is that impacting their business decision? I mean, just kind of generally, how does moving from, I think this is a good value. We bought it last year, we're going to re-up this year to here's all the impressions you got, the quality of the impressions, a score, plus we've uncovered all this additional value. I mean, I would imagine data-driven decision making has got to be so refreshing in these environments. It is, and I think the challenge that a lot of them had was that they were getting the data six to eight weeks later. So if you think of it from a brand perspective, I'm already off to my next sponsorship six to eight weeks later. I can't even think about what I previously did, so for us to be able to give them a solution where they can get their data back in a week or less really helps them make smarter decisions to your point about taking data-driven decision making and figure out real time how they want to adjust to how their audience is adjusting. Do they make a lot of real time corrections in those types of packages or those like annual deals, I would imagine, in the sports thing? Yeah, I think a lot of them at this point are still annual deals the way that they sign up for it, but I think now that they're having access to this data, they're starting to rethink that model and trying to figure out how do we need to change the way that we purchase these things in the future to better fit how they're getting the data around it. Has anyone repriced the inventory based on the data that's come out of the research? You know, they increase the price of a stanchion and decrease, I'm just making stuff up, decrease the price of some other ad unit within the stadium based on some of the data that's come out of your system. We've had a lot of our clients talk about their plans of how they plan to go do that. I think we're only 18 months into this business, so a lot of them are still in the first season or maybe halfway through the second season of working with us, so they're still trying to figure out how to message that properly and what the right channel is for them to recoup those gains, but I think the ability for them to start those conversations is something they've never had before, exposing that to them now allows them to really rethink how their business model has been. It's such a cool example of how data actually allows both halves of the equation to do a better job. I mean, it's really beneficial to everybody, right? It's not just one-sided information that's given somebody a big advantage over the other one. Exactly. All right, so Brian, before I let you go, we're in 2018, it's hard to believe, I can't believe we're almost through the first quarter we've written through it. Some of your priorities for 2018, what is Gum Gum working on? What are you excited about? If we sit down a year from now, what are we going to be talking about? Yeah, I mean, we've been doing this advertising business since 2011, it's our most mature business, so definitely continually scaling that business from an automation standpoint and continually growing that, particularly internationally, has been one of our main goals for this year. As I said, Gum Gum Sports is a pretty new business to us, but we're expecting that to start to bring in significant revenue for us this year and want to see that growth happen. And we're also looking into new emerging areas where we potentially think computer vision can be applied just like we did in Gum Gum Sports. It could be the medical space, it could be television, there's a lot of different applications there that we haven't quite tapped into yet, but we're starting to noodle around what are the right ways that we want to go after that and potentially where we want to invest in with how successful we've been so far. Yeah, well, the exciting opportunity ahead. All right, Brian, he's Brian Kim. He's the Senior Vice President of Product from Gum Gum. Thanks for taking a few minutes out of your day and stopping by. Thanks, Jeff. Pleasure. I'm Jeff Frick. You're watching theCUBE. We'll catch you next time. Thanks for watching.