 Live from the Hilton at Bonnet Creek, Orlando, Florida Extracting the signal from the noise, it's the Cube Covering Vision 2015 Brought to you by IBM And now your hosts Dave Vellante and Jeff Frick Welcome back to IBM Vision everybody I'm Dave Vellante with Jeff Frick This is the Cube, Silicon Angle Goes out to the events We extract the signal from the noise Mark Altschuler is here Vice President of Watson Analytics Mark, good to see you again Nice to see you too So thanks for coming on A little different event here today Smaller but rich Very focused on governance, risk, compliance Sales performance management And the secret sauce The new shiny toy The secret weapon, Watson Analytics So tell us, let's start off Where does Watson Analytics fit In the IBM organization? Yeah, actually great question Get asked this question often So we of course have the Watson Group We just had World of Watson A couple of weeks ago in New York Lots and lots of press around this Dates back to the Jeopardy computer Where we kind of first made our splash Around Watson Watson Analytics is really looking at The structure data aspect So where you have rectangles of data And you need to ingest those You need to statistically interrogate those And kind of find key insights So we're very very focused on The numerical data side Watson Group proper Has a lot of their focus Around the cognitive unstructured data side And then what we do is Or maybe I should back up here How Watson Group views themselves Is they kind of view themselves As the cognitive arms dealer Across IBM We're the first example Within IBM of another IBM solution In another division that has actually consumed A large number of those cognitive services We've put them in Watson Analytics So that you can interact with it in natural language It learns as you go And it has those types of capabilities So they're a strong partner of ours We of course share the name with them They focus more on the unstructured document Ingestion We focus very much on the structure data So at what point did you personally Get involved in Watson? So great question This conference actually marks My three year anniversary in IBM I was part of an acquisition That actually is at this conference now The Verison Acquisition back in May of 2012 I was the co-founder of that company So it was once upon a time Four desks in the basement True basement startup A very fast growing company And actually still here today And it's renamed IBM Sales Performance Management So part of that In terms of this conference is It's really the persona we're targeting Like when we first kind of envisioned this And why we came up with it It's around bringing it to a line of business users Making something that they could consume Something that they could use Something that had the right user experience for them Something that they would understand Be able to share with their colleagues So we're really looking to kind of democratize That skill set around analytics Now we do get into a lot of advanced concepts But we want the users to get there As they're comfortable So we allow them to continue to kind of Peel back the onion Go deeper and deeper as they're comfortable But at any point if they want to pull up and say Oh wow, I've got an interesting insight there Absolutely fine And they'll have that type of experience Okay, so you're an entrepreneur Startup guy And you come into IBM Big company You're a former company Got blue washed Here we go Now you're part of IBM And you see this really cool technology All Watson And you think what I can apply this to analytics And talk a little bit more about it How it came into the show Sure So we were based We had a number of groups Across analytics that we're looking at Data discovery Right, and how did we want to do a data discovery And I've never been a Me Too person The various solutions customers will look at it It was never a Me Too solution It was always about doing something differentiated And we were looking at this But so were the Cognos BI folks So were the SPSS modeler and stats folks We were all looking at What was this next generation Of kind of data discovery And we saw an opportunity There was this opportunity Between data and visualization That we started playing around with That kind of analytic discovery Smart data discovery But this is the idea where You get a piece of data Maybe you know that piece of data But maybe you don't And if you don't Then how do you actually know where to focus And if you do How do you keep your bias out of it Like how often is someone's like CRM pipeline win-loss analysis Why do you win, why do you lose Well you talk to a VP of sales They're going to have their theory They're going to say Well I win when I have a tenured rep Or I win when I'm in this market Or whatever And you end up And I've done this before Making your bias into the visualizations You create visualizations That back up your opinion We looked at this And we said You know what There's this important step In between data visualization Where you statistically Interrogate the data For the users Show them things Either that back up Their assumptions going in Or that strengthen their assumptions Because yeah Rep tenured is probably part of it But it's probably not The complete picture of Why you win, why you lose It's probably a number of factors So we saw that opportunity And we looked across the three groups And we were all looking at it And this was actually a Let's join forces And we're better together Than separate type stories So Watson analytics The true story of Watson analytics Is it's as much about us Disrupting ourselves And it is about us Disrupting the market We had multiple pieces of IP In the labs For many many years Our predict IP Had been sitting in the labs For about four years Our cognitive discovery IP Had been in the labs For three years They were all going to come out As separate products And we said No let's just make them One masterful data discovery Next generation smart data discovery product Okay and it made sense obviously To put that in the analytics group Consuming the cognitive services From Watson And building services And going to market And connectors to our existing technologies We now have a Cognos BI connector Those customers can attach directly The Verison crew on IBM sales performance management Announced their connector here at the conference So they can now send data directly And so we're getting really deeply integrated With all of our products But again you can use it with any product Doesn't have to be our product So I love talking to guys Who started companies, sold them And then stayed with a large company And are innovating inside that large company Conventional wisdom says That the large established players Will get disrupted by the smaller players But increasingly that's not happening The rich get richer in this industry You look around With some exceptions But generally speaking The large established players Seem to have cultivated this innovation culture Through a combination of organic investment And acquisition Is that a fair assessment And what's changed to a lot of those things? I think it's a transformational assessment It's my personal opinion coming into IBM One of the first things I brought to analytics Was you know what We have to do fewer things better Right and we have a partner ecosystem That we can lean on They can bring their solutions to market Always going to be great innovative startups Let them cross the chasm Maybe they'll become interesting technologies To us in the future as well But we have to be willing To transform ourselves If we look at these technologies And when I look at these types of things And kind of get focused on fewer things better I use kind of a VC approach Could I take this idea to a VC And would they fund them And I was looking at a lot of ideas That would be brought forward Where it's like you know what No one would fund this Like it's just I mean you want to build something And then there's an addressable market of five Market's too small or okay right Not scalable, not repeatable Too consulting oriented You can't productize it So now that we start putting this test to things And we look at technologies That have a true barrier to entry Again maybe not permanent But a multi-year head start That allows us to continue To deepen our capabilities as well And the other I think fascinating thing And just to go on a bit of a tangent is One thing I've never done before Like I've done with Watson analytics Is the amount of tooling We have under the cover So we do a lot of observing Around the user experience What are they doing What the experience they're having Where are they clicking Where are they leaving the application But not just where are they Leaving the application Like where are they going to next And so the amount of information We learn just about how people are Losing the product so that we Can tune them continually And this Watson analytics by the way Is a continuous delivery product There's no version 1.0 There's no version 1.1 There's no version 2.0 We just continue to release New functionality all the time So when did you feel that You had the right combination of things To bring together To really have something different It was almost immediate I had actually, I had been sitting In a conference where our data scientists In a box product at the time Called Catalyst was being presented And right away the light bulb went off I didn't have my last roll yet I wasn't over all of the products yet But I was running the performance management product And I'm like, wait a minute This thing it really kind of goes Through all of your data And finds everything that's interesting But it was still a little bit off for me Like it would say something like This isn't a normally shaped bell curve That doesn't really talk to me As a business user So a big part of this was Could we solve taking this stuff That had been historically data scientist oriented And could we move it such that A line of business person Could actually interpret it And I'll tell you one of the funniest stories was We had decided and one of our Marquee visuals is this bullseye spiral And what we do is You put your target in the middle Like I want to predict when loss Is the earlier example I used And then you plot things against the spiral In terms of the strength of How they actually impact the spiral So the closer to the spiral The more predictive it is So your statistical models are going to be closer Although we call it a combination for our users And the key drivers are going to be further away So I sat down with our statistics folks And I described to them this concept How I want to do this circle of influence And plot things against the bullseye And they're like, yeah, yeah, yeah, that's great And I said, so what's the x-axis What's the y-axis And they look at me blankly There's no x and y-axis in stats So just kind of taking this stuff That was the first challenge Is could we take something that historically Gone after a very specific persona Deep data scientist And could we make it consumable by others When we had that breakthrough we knew So I always knew there was something there But that was the kind of crossing Our chasm moment was Could we make something that you and I Would understand with me I mean with either no stats background Or maybe one stats course in college Or something like that I mean that's about the level of knowledge you need And then how important is the cognitive piece The natural language piece In terms of really opening up this As a UI for regular people to use And we often talk in a lot of shows About kind of AG after Google And the expected behavior of things Once Google really got us all spoiled That nothing should be more than a five word question A way that comes back within seconds So now that you guys can start to use That natural language Ask how has that really been a game changer I mean in terms of getting people To their results quickly Like if you think of what was it to do Visualizations before natural language It was grab a data range Choose a chart type, give it a title This is effectively give it a title We'll pick the right set of data We'll default to the best visualization We'll give you backup visualizations that also fit We won't tease you with any visualizations That don't fit unlike if we do that In a spreadsheet where there's a little bit of trial and error Oh that one fits, that one doesn't fit We won't show you a visualization that doesn't fit your data So it completely flipped the paradigm Where it's like you started the title Of what you want to do, we'll fill out the rest You tweak it, you tune it once we do that And then what we also did with it was We moved it into this whole starting points paradigm So even before you ask a question We asked the data in natural language A whole series of questions And we surfaced the most interesting starting points To you before you even start So if it's a new data set to you What are your suggestions? Why don't you start with one of these eight questions? If not, there's always a natural language bar there And you can type in your own question But yeah, it's been transformative to people you can use So how many, I don't know, you said you'd get a lot of Feedback data on the use of the application Typically how many kind of question steps Does somebody go through before they get to the Ah, you know, I'm part of here for a while This is good stuff So it actually depends on what they're doing If they're doing kind of your traditional Kind of data discovery approach Where I've just loaded in a data set And I want to ask my own questions Then they'll see some of our starting points Sometimes those are right on the money But typically they're after something That maybe wasn't the most significant thing in their data set But they still wanted to ask questions about it They'll ask their question about that And usually immediately they ask that first question They see it, then they might drill across They might drill down, they might kind of do A little bit of slicing and dicing on it But usually that first question right away You know, around the NFL playoffs Where we took a whole bunch of sports data And we loaded it in and we let people Just get started with NFL player data And I first started when we were playing with it I started with last year's NFL stats So not this past year, but the season before And one of the questions was Which team had the most pass attempts So if you think two years ago You might think someone like a Denver, right? Peyton Manning was just winging it everywhere And who was a really bad team that year? Cleveland Pass attempts? Cleveland So right away you get something That's really interesting Now on the prediction side You don't really have to know what to ask You pick your target You don't have to think of the question Once you have your targets or multiple targets We give you the profile or pattern of what drives Every target automatically out of the box And we feed that back to you in natural language We give you a visualization Plus a natural language string that explains it Now that's really interesting because The Hadoop movement and the early big data days And the data scientists would always Tell us early on the hardest part Is knowing which questions to ask So you need to be really talented To be able to maybe hack some data Before you can even figure out what questions does Are you saying that you can Actually compress that cycle And tell me what I should be asking? So you just teed up about that If you didn't go to that question I was going to ask it for you So it's a really interesting point that you just stumbled onto So what's the value proposition to a data scientist Let's go to the other extreme Can they use this tool? Is it too rudimentary for them? Will they not get anything interesting? So we had on stage here Jing Shear She's our Chief Statistician IBM Fellow, she was on the main stage keynote And she always So she's the one, she's kind of the brain child behind this The multiple patterns, I think 15 plus patterns That go into this IP She was working on this She set on paper for about 15 years And they coded it for about four years after that So it's absolutely immense IP And to her What she said, where it feeds up for the data scientist Is the data prep cycle So what does she mean by that? She had gone out to one of the municipalities And had met with the mayor there And the mayor wanted to know what drove a certain type What was driving rental rates in different areas of the city What were the true drivers of this And there were also some questions in terms of crime And how it relates to rental rates She spent two weeks with this data set Just trying to figure out what mattered Right? And that's where data scientists Will spend a lot of time is this whole Trial and error of trying to figure out What makes up the model before they can tune the model So for her, she put it in this IP Within two minutes She had that same thing that took her two weeks A few years ago out of the box And so for a data scientist what it does It speeds up that whole find the model process And then you go and tune it With one of the clients we've been We showcased in our professional release Press release Legends Hospitality The stadium operations for the Dallas Cowboys The New York Yankees A few other stadiums They were trying to figure out what drove A certain type of revenue per attendee Stand concessions things like that So he took it And he loaded it into Watts Analytics And right away out of the box he found a model with 82% predictor strength He showed it to his colleague So what was he throwing in? He threw in a whole stadium's information All their concession information He said it was all in a single file But he ingested it in But it was really kind of all the information they Collected so they wouldn't necessarily have Social demographic information other than just Knowing the general nature of who may attend But the more specific kind of stadium related Information they loaded in 82% predictive strength High level model So right away he shows his colleague And his colleague starts laughing and he's like Dude weren't you trying to find that For six months and he starts laughing Now he couldn't finish there right He needed something that was more highly tuned But he could start at that and now he tuned it Using his additional SPSS modeler products And your statistician products And you tuned it further But at that time they were taking it I gave you Jin's example of a two-week example There's a six-month example where they were Just doing so much trial and error to figure out What this high level model was So for every data scientist this is an accelerator We run out of time So we got to let you plug your announcements This week so what are you announcing here At the show? So per usual we do have a free edition That's in the market, it's always been in the market So I encourage anyone go to What'sAnalytics.com Start with the free edition About a month ago we released Or actually a little less than a month ago But mostly being announced here We released our professional edition This is our first enterprise level edition It's the first edition that allows you to do multi-user It has additional data connectors Like the SPM connector We can see you through delivery To add additional connectors In addition to the ones that are there So that's a big announcement for that Professional also allows larger file sizes To be brought into What'sAnalytics And then the other big thing that we announced here Is a grant we've done To our BI customers to get them going Of actual professional licenses And then here at the conference We told our TM1 and SPM customers That we were giving them What'sAnalytics professional For a 12 month term And again, more details people need to follow up With their reps or their advocates But that was probably the biggest announcement We said we're giving it to our customers Excellent, so last question Binoculars, maybe not telescope Put on the binoculars What's this world going to look like In a couple of years I think there's more analytical Capabilities to come We've kind of hung our hat on the cognitive discovery And the predictive so far Coming very shortly our forecasting algorithms So time series based forecasting algorithms So we're very excited about that Scoring capabilities so Maybe you are doing CRM pipeline And you know what the profile of a good lead is So being able to apply that to new leads And get it to the right people when that comes in That's something that's exciting to us So there's a lot of analytic capabilities That are coming in There's a lot of also enterprise and scale Capabilities that are coming in as well A lot of integration with other IBM products Awesome, Mark Altshuler You and your organization applying the Greatness of IBM's cognitive services Into real world examples Thanks so much for coming on theCUBE And sharing that moment Hey thanks so much for having me again Hopefully we'll see you guys at Insight again We'll be there, alright, thank you Keep it right there everybody We'll be back with our next guest This is theCUBE, we're live from IBM Vision 2015 in Orlando We'll be right back