>> Live from New York, It's theCUBE! Covering theCUBE New York City 2018. Brought to you by SiliconANGLE Media and its ecosystem partners. >> Hello and welcome back to theCUBE live in New York City, here for CUBE NYC. In conjunct with Strata Conference, Strata Data, Strata Hadoop This is our ninth year covering the big data ecosystem which has evolved into machine learning, A.I., data science, cloud, a lot of great things happening all things data, impacting all businesses I'm John Furrier, your host with Dave Vellante and Peter Burris, Peter is filling in for Dave Vellante. Next guest, Stephanie McReynolds who is the CMO, VP of Marketing for Alation, thanks for joining us. >> Thanks for having me. >> Good to see you. So you guys have a pretty spectacular exhibit here in New York. I want to get to that right away, top story is Attack of the Bots. And you're showing a great demo. Explain what you guys are doing in the show. >> Yah, well it's robot fighting time in our booth, so we brought a little fun to the show floor my kids are.. >> You mean big data is not fun enough? >> Well big data is pretty fun but occasionally you got to get your geek battle on there so we're having fun with robots but I think the real story in the Alation booth is about the product and how machine learning data catalogs are helping a whole variety of users in the organization everything from improving analyst productivity and even some business user productivity of data to then really supporting data scientists in their work by helping them to distribute their data products through a data catalog. >> You guys are one of the new guard companies that are doing things that make it really easy for people who want to use data, practitioners that the average data citizen has been called, or people who want productivity. Not necessarily the hardcore, setting up clusters, really kind of like the big data user. What's that market look like right now, has it met your expectations, how's business, what's the update? >> Yah, I think we have a strong perspective that for us to close the final mile and get to real value out of the data, it's a human challenge, there's a trust gap with managers. Today on stage over at STRATA it was interesting because Google had a speaker and it wasn't their chief data officer it was their chief decision scientist and I think that reflects what that final mile is is that making decisions and it's the trust gap that managers have with data because they don't know how the insides are coming to them, what are all the details underneath. In order to be able to trust decisions you have to understand who processed the data, what decision making criteria did they use, was this data governed well, are we introducing some bias into our algorithms, and can that be controlled? And so Alation becomes a platform for supporting getting answers to those issues. And then there's plenty of other companies that are optimizing the performance of those QUERYS and the storage of that data, but we're trying to really to close that trust gap. >> It's very interesting because from a management standpoint we're trying to do more evidence based management. So there's a major trend in board rooms, and executive offices to try to find ways to acculturate the executive team to using data, evidence based management healthcare now being applied to a lot of other domains. We've also historically had a situation where the people who focused or worked with the data was a relatively small coterie of individuals that crave these crazy systems to try to bring those two together. It sounds like what you're doing, and I really like the idea of the data scientists, being able to create data products that then can be distributed. It sounds like you're trying to look at data as an asset to be created, to be distributed so they can be more easily used by more people in your organization, have we got that right? >> Absolutely. So we're now seeing we're in just over a hundred production implementations of Alation, at large enterprises, and we're now seeing those production implementations get into the thousands of users. So this is going beyond those data specialists. Beyond the unicorn data scientists that understand the systems and math and technology. >> And business. >> And business, right. In business. So what we're seeing now is that a data catalog can be a point of collaboration across those different audiences in an enterprise. So whereas three years ago some of our initial customers kept the data catalog implementations small, right. They were getting access to the specialists to this catalog and asked them to certify data assets for others, what were starting to see is a proliferation of creation of self service data assets, a certification process that now is enterprise-wide, and thousands of users in these organizations. So Ebay has over a thousand weekly logins, Munich Reinsurance was on stage yesterday, their head of data engineering said they have 2,000 users on Alation at this point on their data lake, Fiserv is going to speak on Thursday and they're getting up to those numbers as well, so we see some really solid organizations that are solving medical, pharmaceutical issues, right, the largest re insurer in the world leading tech companies, starting to adopt a data catalog as a foundation for how their going to make those data driven decisions in the organization. >> Talk about how the product works because essentially you're bringing kind of the decision scientists, for lack of a better word, and productivity worker, almost like a business office suite concept, as a SAS, so you got a SAS model that says "Hey you want to play with data, use it but you have to do some front end work." Take us through how you guys roll out the platform, how are your customers consuming the service, take us through the engagement with customers. >> I think for customers, the most interesting part of this product is that it displays itself as an application that anyone can use, right? So there's a super familiar search interface that, rather than bringing back webpages, allows you to search for data assets in your organization. If you want more information on that data asset you click on those search results and you can see all of the information of how that data has been used in the organization, as well as the technical details and the technical metadata. And I think what's even more powerful is we actually have a recommendation engine that recommends data assets to the user. And that can be plugged into Tablo and Salesworth, Einstein Analytics, and a whole variety of other data science tools like Data Haiku that you might be using in your organization. So this looks like a very easy to use application that folks are familiar with that you just need a web browser to access, but on the backend, the hard work that's happening is the automation that we do with the platform. So by going out and crawling these source systems and looking at not just the technical descriptions of data, the metadata that exists, but then being able to understand by parsing the sequel weblogs, how that data is actually being used in the organization. We call it behavior I.O. by looking at the behaviors of how that data's being used, from those logs, we can actually give you a really good sense of how that data should be used in the future or where you might have gaps in governing that data or how you might want to reorient your storage or compute infrastructure to support the type of analytics that are actually being executed by real humans in your organization. And that's eye opening to a lot of I.T. sources. >> So you're providing insights to the data usage so that the business could get optimized for whether it's I.T. footprint component, or kinds of use cases, is that kind of how it's working? >> So what's interesting is the optimization actually happens in a pretty automated way, because we can make recommendations to those consumers of data of how they want to navigate the system. Kind of like Google makes recommendations as you browse the web, right? >> If you misspell something, "Oh did you mean this", kind of thing? >> "Did you mean this, might you also be interested in this", right? It's kind of a cross between Google and Amazon. Others like you may have used these other data assets in the past to determine revenue for that particular region, have you thought about using this filter, have you thought about using this join, did you know that you're trying to do analysis that maybe the sales ops guy has already done, and here's the certified report, why don't you just start with that? We're seeing a lot of reuse in organizations, wherein the past I think as an industry when Tablo and Click and all these B.I tools that were very self service oriented started to take off it was all about democratizing visualization by letting every user do their own thing and now we're realizing to get speed and accuracy and efficiency and effectiveness maybe there's more reuse of the work we've already done in existing data assets and by recommending those and expanding the data literacy around the interpretation of those, you might actually close this trust gap with the data. >> But there's one really important point that you raised, and I want to come back to it, and that is this notion of bias. So you know, Alation knows something about the data, knows a lot about the metadata, so therefore, I don't want to say understands, but it's capable of categorizing data in that way. And you're also able to look at the usage of that data by parsing some of sequel statements and then making a determination of the data as it's identified is appropriately being used based on how people are actually applying it so you can identify potential bias or potential misuse or whatever else it might be. That is an incredibly important thing. As you know John, we had an event last night and one of the things that popped up is how do you deal with emergence in data science in A.I, etc. And what methods do you put in place to actually ensure that the governance model can be extended to understand how those things are potentially in a very soft way, corrupting the use of the data. So could you spend a little bit more time talking about that because it's something a lot of people are interested in, quite frankly we don't know about a lot of tools that are doing that kind of work right now. It's an important point. >> I think the traditional viewpoint was if we just can manage the data we will be able to have a govern system. So if we control the inputs then well have a safe environment, and that was kind of like the classic single source of truth, data warehouse type model. >> Stewards of the data. >> What we're seeing is with the proliferation of sources of data and how quickly with IOT and new modern sources, data is getting created, you're not able to manage data at that point of that entry point. And it's not just about systems, it's about individuals that go on the web and find a dataset and then load it into a corporate database, right? Or you merge an Excel file with something that in a database. And so I think what we see happening, not only when you look at bias but if you look at some of the new regulations like [Inaudible] >> Sure. Ownership, [Inaudible] >> The logic that you're using to process that data, the algorithm itself can be biased, if you have a biased training data site that you feed it into a machine learning algorithm, the algorithm itself is going to be biased. And so the control point in this world where data is proliferating and we're not sure we can control that entirely, becomes the logic embedded in the algorithm. Even if that's a simple sequel statement that's feeding a report. And so Alation is able to introspect that sequel and highlight that maybe there is bias at work and how this algorithm is composed. So with GDPR the consumer owns their own data, if they want to pull it out from a training data set, you got to rerun that algorithm without that consumer data and that's your control point then going forward for the organization on different governance issues that pop up. >> Talk about the psychology of the user base because one of the things that shifted in the data world is a few stewards of data managed everything, now you've got a model where literally thousands of people of an organization could be users, productivity users, so you get a social component in here that people know who's doing data work, which in a way, creates a new persona or class of worker. A non techy worker. >> Yeah. It's interesting if you think about moving access to the data and moving the individuals that are creating algorithms out to a broader user group, what's important, you have to make sure that you're educating and training and sharing knowledge with that democratized audience, right? And to be able to do that you kind of want to work with human psychology, right? You want to be able to give people guidance in the course of their work rather than have them memorize a set of rules and try to remember to apply those. If you had a specialist group you can kind of control and force them to memorize and then apply, the more modern approach is to say "look, with some of these machine learning techniques that we have, why don't we make a recommendation." What you're going to do is introduce bias into that calculation. >> And we're capturing that information as you use the data. >> Well were also making a recommendation to say "Hey do you know you're doing this? Maybe you don't want to do that." Most people are using the data are not bad actors. They just can't remember all the rule sets to apply. So what were trying to do is cut someone behaviorally in the act before they make that mistake and say hey just a bit of a reminder, a bit of a coaching moment, did you know what you're doing? Maybe you can think of another approach to this. And we've found that many organizations that changes the discussion around data governance. It's no longer this top down constraint to finding insight, which frustrates an audience, is trying to use that data. It's more like a coach helping you improve and then social aspect of wanting to contribute to the system comes into play and people start communicating, collaborating, the platform and curating information a little bit. >> I remember when Microsoft Excel came out, the spreadsheet, or Lotus 123, oh my God, people are going to use these amazing things with spreadsheets, they did. You're taking a similar approach with analytics, much bigger surface area of work to kind of attack from a data perspective, but in a way kind of the same kind of concept, put the hands of the users, have the data in their hands so to speak. >> Yeah, enable everyone to make data driven decisions. But make sure that they're interpreting that data in the right way, right? Give them enough guidance, don't let them just kind of attack the wild west and fair it out. >> Well looking back at the Microsoft Excel spreadsheet example, I remember when a finance department would send a formatted spreadsheet with all the rules for how to use it out of 50 different groups around the world, and everyone figured out that you can go in and manipulate the macros and deliver any results they want. And so it's that same notion, you have to know something about that, but this site, in many respects Stephanie you're describing a data governance model that really is more truly governance, that if we think about a data asset it's how do we mediate a lot of different claims against that set of data so that its used appropriately, so its not corrupted, so that it doesn't effect other people, but very importantly so that the out6comes are easier to agree upon because there's some trust and there's some valid behaviors and there's some verification in the flow of the data utilization. >> And where we give voice to a number of different constituencies. Because business opinions from different departments can run slightly counter to one another. There can be friction in how to use particular data assets in the business depending on the lens that you have in that business and so what were trying to do is surface those different perspectives, give them voice, allow those constituencies to work that out in a platform that captures that debate, captures that knowledge, makes that debate a knowledge of foundation to build upon so in many ways its kind of like the scientific method, right? As a scientist I publish a paper. >> Get peer reviewed. >> Get peer reviewed, let other people weigh in. >> And it becomes part of the canon of knowledge. >> And it becomes part of the canon. And in the scientific community over the last several years you see that folks are publishing their data sets out publicly, why can't an enterprise do the same thing internally for different business groups internally. Take the same approach. Allow others to weigh in. It gets them better insights and it gets them more trust in that foundation. >> You get collective intelligence from the user base to help come in and make the data smarter and sharper. >> Yeah and have reusable assets that you can then build upon to find the higher level insights. Don't run the same report that a hundred people in the organization have already run. >> So the final question for you. As you guys are emerging, starting to do really well, you have a unique approach, honestly we think it fits in kind of the new guard of analytics, a productivity worker with data, which is we think is going to be a huge persona, where are you guys winning, and why are you winning with your customer base? What are some things that are resonating as you go in and engage with prospects and customers and existing customers? What are they attracted to, what are they like, and why are you beating the competition in your sales and opportunities? >> I think this concept of a more agile, grassroots approach to data governance is a breath of fresh air for anyone who spend their career in the data space. Were at a turning point in industry where you're now seeing chief decision scientists, chief data officers, chief analytic officers take a leadership role in organizations. Munich Reinsurance is using their data team to actually invest and hold new arms of their business. That's how they're pushing the envelope on leadership in the insurance space and were seeing that across our install base. Alation becomes this knowledge repository for all of those mines in the organization, and encourages a community to be built around data and insightful questions of data. And in that way the whole organization raises to the next level and I think its that vision of what can be created internally, how we can move away from just claiming that were a big data organization and really starting to see the impact of how new business models can be creative in these data assets, that's exciting to our customer base. >> Well congratulations. A hot start up. Alation here on theCUBE in New York City for cubeNYC. Changing the game on analytics, bringing a breath of fresh air to hands of the users. A new persona developing. Congratulations, great to have you. Stephanie McReynolds. Its the cube. Stay with us for more live coverage, day one of two days live in New York City. We'll be right back.

(stirring music) >> Hi, I'm Peter Burris, and welcome to another CUBE Conversation from theCUBE Studios in beautiful Palo Alto, California. Got a great conversation today. We're going to be talking about some of the new advances that are associated with big data analytics and improving the rate at which human beings, people who actually work with data, can get more out of their data, be more certain about their data, and improve the social system that actually is dependent upon data. To do that, we've got Aaron Kalb of Alation here with us. Aaron is the co-founder and is VP of design and strategic initiatives. Aaron, welcome back to theCUBE. >> Thanks so much for having me, Peter. >> So, then, let's start this off. The concern that a lot of folks have when they think about analytics, big data, and the promise of some of these new advanced technologies is they see how they could be generating significant business value, but they observe that it often falls short. It falls short for technological reasons, you know, setting up the infrastructure is very, very, difficult. But we've started solving that by moving a lot of these workloads to the cloud. They also are discovering that the toolchains can be very complex, but they're starting to solve that by working with companies with vision, like Alation, about how you can bring these things together more easily. There are some good things happening within the analytics space, but one of the biggest challenges is, even if you set up your pipelines and your analytics systems and applications right, you still encounter resistance inside the business, because human beings don't necessarily have a natural affinity for data. Data is not something that's easy to consume, it's not something easy to recognize. People just haven't been trained in it. We need more that makes it easy to identify data quality, data issues, et cetera. Tell us a little bit about what Alation's doing to solve that human side, the adoption side of the challenge. >> That's a great point and a great question, Peter. Fundamentally, what we see is it used to be a problem of quantity. There wasn't enough ability to generate data assets, and to distribute them, and to get to them. Now, there's just an overwhelming amount of places to gather data. The problem becomes finding development data for your need, understanding and putting it into context, and most fundamentally, trusting that it's actually telling you a true story about the world. You know, what we find now is, as there's been more self-service analytics, there's more and more dashboards and queries and content being generated, and often an executive will look at two different answers to the same question that are trending in totally different directions. They'll say, "I can't trust any of this. "On paper, I want to be data-driven, "but in actuality, I'm just going to go back to my gut, "'cause the data is not always trustworthy, "and it's hard to tell what's trustworthy and what's not." >> This is, even after they've found the data and enough people have been working on it to say, to put it in context to say, "Yes, this data is being used in marketing," or, "This data has been used in operations production." there's another layer of branding or whatnot that we can put on data that says, "This data is appropriate for use in this way." Is that what we're talking about here? >> Absolutely right. To help with finding and understanding data, you can group it and make it browsable by topic. You can enable keyword search over it in that natural language. That's stuff that Alation has done in the past. What we're excited to unveil now is this idea of trust check, which is all about saying, wherever you're at in that data value chain of taking raw data and schematizing it and eventually producing pretty dashboards and visualizations, that at every step, we can ensure that only the most trustworthy data sets are being used, because any problem upstream flows downstream. >> So, trust check. >> Trust check. >> Trust check, it's something that comes out of Alation. Is it also being used with other visualization tools or other sources or other applications? >> That's a great question. It's all of the above. Trust check starts with saying, if I'm an analyst who wants to create a dashboard or a visualization, I'm going to have to write some SQL query to do that. What we've done in that context with Alation Compose, is our home-grown SQL tool, is provided a tool, and trust check kind of gets its name from spell check. It used to be there was a dictionary, and you could look it up by hand, and you could look it up online, but that's a lot of work for every single word to check it. And then, you know, Microsoft, I think, was the first innovative saying, "Oh, let's put a little red squiggle that you can't miss "right in your workflow as you're writing, "so you don't have to go to it, it comes to you." We do the exact same thing. I'm about to query a table that is deprecated or has a data quality issue. I immediately see bright red on my screen, can't miss it, and I can fix my behavior. That's as I'm creating a data asset. We also, through our partnerships with Salesforce and with Tableau, each of whom have very popular visualization tools, to, say. if people are consuming a dashboard, not a SQL query, but looking at a Tableau dashboard or a visualization in Salesforce Einstein Analytics, what would it mean to badge right there and then, put a stamp of approval on the most trustworthy sources and a warning or caveat on things that might have an upstream data quality problem? >> So, when you say warning or caveat, you're saying literally that there are exceptions or there are other concerns associated with the data, and reviewing that as part of the analytic process. >> That's exactly right. Much like, again, spell check underlines, or looking at, if you think about if I'm driving in my car with Waze, and it says, "Oh, traffic up ahead, view route this way." What does it mean to get in the user interface where people live, whether they're a business user in Salesforce or Tableau, or a data analyst in a query tool, right there in their flow having onscreen indications of everything happening below the tip of the iceberg that affects their work and the trustworthiness of the data sets they're using. >> So that's what it is. I'll tell you a quick story about spell check. >> Please. >> Many years ago, I'm old enough that I was one of the first users of some of these tools. When you typed in IBM, Microsoft Word would often change it to DUM, which was kind of interesting, given the things that were going on between them. But it leads you to ask questions. How does this work? I mean, how does spell check work? Well, how does trust check work, because that's going to have an enormous implication. People have to trust how trust check works. Tell us a little bit about how trust check works. >> Absolutely. How do you trust trust check? The little red or yellow or bright, salient indicators we've designed are just to get your attention. Then, as a user, you can click into those indicators and see why is this appearing. The biggest reason that an indicator will appear in a trust check context is that a person, a data curator or data steward, has put a warning or a deprecation on the data set. It's not, you know, oh, IBM doesn't like Microsoft, or vice versa. You know, you can see the sourcing. It isn't just, oh, because Merriam-Webster says so. It emerges from the logic of your own organization. But now Alation has this entire catalog backing trust check where it gives a bunch of signals that can help those curators and stewards to decide what indicators to put on what objects. For example, we might observe, this table used to be refreshed frequently. It hasn't in a while. Does that mean it's ripe for getting a bit of a warning on it? Or, people aren't really using this data set. Is there a reason for that? Or, something upstream was just flagged having a data quality issue. That data quality issue might flow downstream like pollution in a creek, and that can be an indication of another reason why you might want to label data as not trustworthy. >> In Alation context with Salesforce and Tableau partners, and perhaps some others, this trust check ends up being a social moniker for what constitutes good data that is branded as a consequence of both technological as well as social activities around that data captured by Alation. I got that right? >> That's exactly right. We're taking technical signals and social signals, because what happens in our customers today before we launched trust check, what they would do is, if you had the time, you would phone a friend. You'd say, "Hey, you seem to be data-savvy. "Does this number look weird to you? "Do you know what's going on? "Is something wrong with the table that it's sourced from?" The problem is, that person's on vacation, and you're out of luck. This is saying, let's push everything we know across that entire chain, from the rawest data to the most polished asset and have all that information pushed up to where you live in the moment you're making a decision, should I trust this data, how should I use it? >> In the whole, going back to this whole world of big data and analytics, we're moving more of the workloads to the cloud to get rid of the infrastructure problems. We're utilizing more integrated toolchains to get rid of the complexity associated with a lot of the analytic pipelines. How does trust check then applied, go back to this notion of human beings not being willing to accept somebody else's data. Give us that use case of how someone's going to sit down in a boardroom or at a strategic meeting or whatever else it is, see trust check, and go, "I get it." >> Absolutely, that's a fantastic question. There's two reasons why, even though all organizations, or 80% according to Gartner, claim they're committed to being data-driven. You still have these moments, people say, "Yeah, I see the numbers, "but I'm going to ignore them, or discount them, "or be very skeptical of them." One issue is just how much of the data that gets to you in the boardroom or the exec team meeting is wrong. We had an incredibly successful data-driven customer who did an internal audit and found that 1/3 of the numbers that appeared in the PowerPoint presentations on which major business decisions were being made, a full 1/3 of them were off by an extraordinary amount, an amount so big that it would, the decision would've cut the other way had the number been accurate. The sheer volume of bad data coming in to undermine trust. The second is, even if only 5% of the data were untrustworthy, if you don't know which is which, the 95% that's trustworthy and the 5% that's not, you still might not be able to use it with confidence. We believe that having trust check be at every stage in this data value chain will solve, actually, both problems by having that spell-check-like experience in the query tool, which is where most analytics projects start. We can reduce the amount of garbage going into the meeting rooms where business choices are being made. And by putting that badge saying "This is certified," or, "Take this with a grain of salt," or, "No, this is totally wrong," that putting that badge on the visualizations that business leaders are looking at in Salesforce and Tableau, and over time, in ideally every tool that anybody would use in an enterprise, we can also help distinguish the wheat from the chaff in that context as well. We think we're attacking both parts of this problem, and that will really drive a data-driven culture truly being adoptable in an organization. >> I want to tie a couple things that you said here. You mentioned the word design a couple times. You're the VP of design at Alation. It also sounds like when you're talking about design, you're not just talking about design of the interface or the software. You're talking about design of how people are going to use the software. What is the extent to which design, what's the scope of design as you see it in this context of advanced analytics, and is trust check just a first step that you're taking? Tell us a little bit about that. >> Yeah, that's a great set of questions, Peter. Design for us means really looking at humans, and starting by listening and watching. You know, a lot of people in the cataloging space and the governance space, they list a lot of should statements. "People should adopt this process, "because otherwise, mistakes will be made." >> Because Gartner said 80% of you have! >> Right, exactly. We think the shoulds only get you so far. We want to really understand the human psychology. How do people actually behave when they're under pressure to move quickly in a rapidly changing environment, when they're afraid of being caught having made a mistake? There's all these pressures people are under. And so, it's not realistic to say, again, you could imagine saying, "Oh, every time before you go out the door, "go to MapQuest or some sort of traffic website "and look up the route and print it out, "so you make sure you plot correctly." No one has time for that, just like no one has time to look up every single word in their essay or their memo or their email and look it up in the dictionary to see if it's right. But when you have an intervention that comes into somebody's flow and is impossible to miss, and is an angel on your shoulder keeping you from making a mistake, or, you know, in-car navigation that tells you in real time, "Here's how you should route." Those sort of things fit into somebody's lifestyle and actually move impact. Our idea is, let's meet people where they are. Acknowledge the challenges that humans face and make technology that really helps them and comes to them instead of scolding them and saying, "Oh, you should change your flow in this uncomfortable way "and come to us, "and that's the only way "you'll achieve the outcome you want." >> Invest the tool into the process and into the activity, as opposed to force people to alter the activity around the limitations or capabilities of the tool. >> Exactly right. And so, while design is optimizing the exact color and size and UI/UX both in our own tools and working with our partners to optimize that, it's starting at an even bigger level of saying, "How do we design the entire workflow "so humans can do what they do best "and the computer just gives them "what they need in real time?" >> And as something as important, and this kind of takes it full circle, something as important and potentially strategic as advanced analytics, having that holistic view is really going to determine success or failure in a lot of businesses. >> That is absolutely right, Peter, and you asked earlier, "Is this just the beginning?" That's absolutely true. Our goal is to say, whatever part of the analytics process you are in, that you get these realtime interventions to help you get the information that's relevant to you, understand what it means in the context you're in, and make sure that it's trustworthy and reliable so people can be truly data-driven. >> Well, there's a lot of invention going on, but what we're really seeking here is changes in social behavior that lead to consequential improvements in business. Aaron Kalb, VP of design and strategic initiatives at Alation, thanks very much for talking about this important advance in how we think about analytics. >> Thank you so much for having me, Peter. >> This is, again, Peter Burris. This has been a CUBE Conversation. Until next time. (stirring music)