 Live from Boston, Massachusetts, it's theCUBE, covering Spark Summit East 2017, brought to you by Databricks. Now, here are your hosts, Dave Vellante and George Gilbert. Welcome back to Boston, everybody. This is theCUBE. We're live here at Spark Summit East in the Heinz Convention Center. This is theCUBE. Check out siliconangle.com for all the news of the day. Check out wikibon.com for all the research. I'm really excited about this session here. ALS is here as the Vice President of Analytics and R&D at McGraw Hill Education. And I'm so excited because we always talk about digital transformations and transformations. We have an example of a 150-year-old company that has been, I'm sure, to many transformations. We're going to talk about a recent one. ALS, a welcome to theCUBE. Thanks for coming on. Thank you, a pleasure to be here. So you heard my little narrative up front. You obviously have not been with the company for 150 years. I talked about all the transformations, but there's certainly one that's recent in the last couple of years anyway, which is digital. We know McGraw Hill is a print publisher. Describe your business. Yeah, so McGraw Hill Education has been traditionally a print publisher, but beginning with our new CEO, David Levin. He joined the company about two years ago. And now we call ourselves a learning science company. So it's no longer print polishing. It's smart digital, and by smart digital, we mean we're trying to transform education by applying principles of learning science. And basically what that means is, we try to understand how do people learn and how they can learn better. And so there are a number of domains, cognitive science, brain sciences, data science, and we begin to try to understand what are the known knowns in these areas and then apply it to education. I think Mark Benioff said at first, at least the first I heard, he said there can be way more SaaS companies that come out of non-tech companies than tech companies. We're talking off camera, you're a software company. Describe that in some detail. Yeah, so being a software company is new for us, but we've moved pretty quickly. Our core competency has been really expert knowledge about education. We work with educators, subject matter experts. So for over 100 years, we've created vetted content assessments and so on. So we have a great deal of domain expertise in education. And now we're taking sort of the new area of frontiers of knowledge and cognitive science, brain sciences. How can learners learn better in applying that to software and models and algorithms? Okay, and there's a data component to this as well. So yeah, the way I think about it is we're a smart digital company, but smart digital is fueled by smart data. It's data is underlies everything that we do. Why? Because in order to strengthen learners, provide them with the optimal pathway, as well as instructors. We believe instructors are at the center of this new transformation. We need to provide immediate real-time data to students and instructors on how am I doing? How can I do better? This is the predictive component. And then you're telling me maybe I'm not on the best path. So what's my, how can I do better? The optimal path. And so all of that is based on data. Okay, and so that's, I mean, the master question. Do you do any print anymore? Yes, we still do print because there's still a huge need for print. So print's not going to go away, right? Okay, I just wanted to clarify that. But what you described is largely a business model change. Not largely, it is a business model change. But also the value proposition is changing. You're providing a new service related, but new incremental value, right? Yeah, so the value proposition has changed. And here again, data is critical. Inquiring minds want to know, our customers want to know. All right, we're going to use your technology and your products and solutions show us rigorously, empirically that it works, right? That's the bottom line question. Is it effective? Are the tools, product solutions, not just ours, but our products and solutions have a context. So is the instruction effective? Is it effective for everyone, right? So all that is relying on data. So how much of a course, how much of the content in a course would you prepare? Is it now the entire course where, and you instrument the student's interaction with it, and then essentially you're selling the outcomes, the improved outcomes? Yeah, I think that's one way to think about it. Here's another model change. So this is not so much digital versus non-digital, but we've been a closed environment. You buy a textbook from us, all the material, the assessments is McGraw-Hill Education. But now a fundamental part of our thinking as a software company is that we have to be an open company. Doesn't mean open as in free, but it's an open ecosystem. So one of the things that we believe in very much is standards. So there's a standard body in education called IMS Global. My boss, Stephen Laster, is on the board of IMS Global. So think of that as, this encompasses everything from different tools, working together, interoperability tools, or interoperability standards, data standards for data exchange. So we will always produce great content, great assessments. We have amazing platform and analytics capability. However, we don't believe all of our customers are going to want to use everything from McGraw-Hill. So interoperability standards, data standards is vital to what we're doing. Can you explain in some detail this learning science company, explain how we learn? You were talking off camera about sort of the three... Yeah, so this is just one example. It's well known that memory decays exponentially, meaning when you see some item of knowledge for the first time, unless something happens, it goes into short-term memory and then it evaporates. So one of the challenges in education is, how can I acquire knowledge and retain knowledge? Now most of the techniques that we all use are not optimal. We cram right before an exam, we highlight things, and that creates the illusion that we'll be able to recall it, but it's an illusion. Now cognitive science and research in cognitive science tells us that there are optimal strategies for acquiring knowledge and recalling it. So three examples of that are effort for recall. If you have to actively recall some item of knowledge, that helps with the stickiness. Another is space practice, practicing out your recall over multiple sessions, and then another one is interleaving. So what we do is, we just recently came out of the product last week called StudyWise, and what we've done is taken those principles, written some algorithms, applied those algorithms into a mobile product, and so that's going to allow learners to optimize their acquisition and recall knowledge. And you're using Spark, too. Yeah, we're using Spark and we're using Databricks. So we use, so I think what's important there is, not just Spark is a technology, but it's an ecosystem. It's a set of technologies, and it has to be woven together into a workflow. So everything from building the model and algorithm, and those are always first approximations. We do the best that we can in terms of how we think the algorithm should work, and then deploy that. So our data science team and learning science team builds the models, designs the models, but our IT team wants to make sure that it's part of a workflow. They don't want to have to deal with a new set of technologies. So essentially pressing the button goes into production, and then it doesn't stop there because as StudyWise has gone to the market last week, now we're collecting data real time as learners are interacting with our products, the results of their interactions is coming into our research environment, and we're analyzing that data as a way of updating our models and tuning the models. So would it be fair to say that, it was interesting when you talked about these new ways of learning, if I were to create an analogy to legacy enterprise apps, they standardize business transactions on the workflows that went with them. It's like you're picking out the best practices in learning, codifying them into an application, and you've made it, you've opened it up so other platforms can take some or all, and then you're taking live feedback from the models, but not just tuning the existing model, but actually adding learnings to the model over time as you get a better sense for how effort of recall works or interleaving works. Yeah, I think it's exactly right of, I do want to emphasize something, an aspect of what you said is, we believe, and it's not just we believe, the research in Learning Science shows that we can get the best, the most significant learning gains when we place the instructor, the master teacher at the center of learning, right? And doing that, not just in isolation, but what we want to do is create a community of practitioners, master teachers. So think of the healthcare analogy. We have expert physicians. So when we have a new technique or even an old technique, what's working, what's not working? Let's look at the data. So what we're also doing is instrumenting our tools so that we can surface these insights to the master practitioners or master teachers. George is trying this technique that's working or not working. What adjustments do we need to make? So it's not just something has to happen with the learner. Maybe we need to adjust our curriculum. I have to change my teaching practices, my assessments. And the incentive for the master practitioners to collaborate is because that's just their nature? I think it is. So let's kind of stand back. I think the current paradigm of instruction is lecture mode. I want to impart knowledge. So I'm going to give a lecture. And then assessment is time tests. In the educational jargon for that is summative assessments. So lecture and tests. That's the dominant paradigm in education. All the research evidence says that doesn't work. It doesn't work, but we still do it. For how many hundreds of years? Well, it was okay if we needed to train and educate a handful of people. But now everyone needs to be educated and it's lifelong learning. So that paradigm doesn't work. And the research evidence is overwhelming that it doesn't work. So we have to change our paradigm where the new paradigm, and this is again based on research, is differentiated instruction. So different learners are at different stages and that are learning. And depending on what you need to know, I'm at a different stage. So we need assessments. So assessments are not punitive, they're not tests. They help us determine what kind of knowledge, what kind of information each learner needs to know. And the instructor helps with the differentiated instruction. It's an alignment. It's an alignment, yeah. And then really to take it to the next stage, the master practitioners, if they are armed with the right data, they can begin to compare. All right, practices, this way of teaching for these types of students works well. These are the adjustments that we need to make. So bringing it down to Earth with Spark, these models of how to teach or perhaps how to differentiate the instruction, how to do differentiated assessments, these are the Spark models. Yeah, these are the Spark models. So let's kind of stand back and see what's different about traditional analytics or business intelligence and the new analytics enabled by Spark and so on. So first, traditional analytics, the questions that you need to be able to answer are defined beforehand. And then they're implemented in schemas in a data warehouse. In the new order of things, I have questions that I need to ask and they just arise right now. So I'm not going to anticipate all the questions that I'm going to be able to ask. So we have to be enable the ability to ask new questions and be able to receive answers immediately. Second, the feedback loop. Traditional analytics is batch mode, right? Overnight data warehouse gets updated. Well, imagine you're flying an airplane, you're the pilot, a new weather system emerges. You can't wait a week or six months to get a report. I have to have corrective course. I have to re-navigate and find a new course. So same way, a student encounters difficulty. Tell me what I need to do, what course correction do I need to apply? So the data has to come in real time, the models have to run real time, and if it's at scale, then we have to have parallel processing, and then the updates, the round trip data back to the instructor or the student has to be essentially real time or near real time. Spark is one of the technologies that's enabling that. The way you got here is kind of interesting. You used to be CIO. All right, got that big Yale brain working for you. You're not a developer, I presume, is that right? Or how did you end up in this role? I think it's really a passion for education, and I think this is at McGraw Hill. So I'm a first generation college student. I feel myself really, I went to public school in Los Angeles. I had a lot of great breaks. I had great teachers who inspired me. And so I think first it's education, but I think we have a major, major problem that we need to solve, right? So if we look at, so I spent five years with the Minnesota State Colleges and University System. Most of the colleges, community colleges are open access institutions. So let me just give you a quick statistic. 70% of students who enter community colleges are not prepared in math and English. So seven out of 10 students need remediation. Of the seven out of 10 students who need remediation, only 15%, not 5.0, one five succeed to the next level. So this is a national tragedy. And that's at the community college level. That's at the community college level. So we're talking about millions of students who are not making it past the first gate. And they go away thinking they failed, they incurred debt. And their life is now stuck, okay? So this is playing itself out not to tens of thousands of students, but hundreds of thousands of students annually, right? So we've got to solve this problem, right? Of, and I think, so it's not technology but reshaping the paradigm of how we think about education. It is a national disaster because oftentimes that's the only affordable route for folks and they are taking on debt, thinking, okay, this is the gateway. Al, we have to leave it there. Awesome segment. Thanks very much for coming to theCUBE really for you. Thank you very much. All right, you're welcome. Keep it right there, buddy. George and I will be back with our next guest. This is theCUBE. We're live from Boston. We're right back.