Jonas Christensen 2:39
David Mariani, welcome to Leaders of Analytics. It is fantastic to have you on the show.
David Mariani 2:47
Thanks for inviting me, Jonas.
Jonas Christensen 2:49
It is my pleasure to have you on here. And we are going to have a very interesting conversation today about how to treat your data as a strategic asset and all the tech that drives data science. So I for one, I'm very interested and excited to learn from you. I've looked at your resume and your experience, and there is so much to learn today. So, let's get straight into that. But before we get to all the technical detail, we'd love to learn more about you, Dave. So in your own words, could you tell us a bit about yourself, your career background and what you do?
David Mariani 3:21
Yeah sure, Jonas. So, my background is - for a tech person and a startup founder - It's not typical. So, I grew up in a farm town in central California. And you know, of parents who never went to college. So I taught myself basic programming, using on an Apple II Plus in high school, and thought it was the coolest thing ever. But of course, nobody told me that there was this industry called software engineering. So I chose UCLA as my college, because it looked like what a college should look like to me. No other reason other than that. And I majored in economics, because that was the closest thing to business, which is what I thought I wanted to do. So having no clue, you know, basically that the technology industry really existed. When I interviewed out of school, I ran into this funny company that was a consulting company that did consulting for risk management for banks. I love the people. It was just a handful of people - 10 people. So it was a startup in terms of size, but it also became a startup in terms of software. So the reason why they hired me was because I knew how to program basic and they wanted me to write dBase reports, which I had no clue what dBase was, but quickly learned it. So long story short, it was for me coming from the business side, but then knowing and learning technology and learning software, and how to write code and how to do it in a company. We morphed that company into a software company from a consulting firm and then sold it to Oracle for over $120 million during the day. And then, that sort of launched me off on this career of really a tech startup before they were called tech startups. So I learned a lot from that experience. And having sort of - I call ''living the problem'' - saw a lot of our financial clients having real trouble with the reporting. And the reporting during the day was the sort of banded report writers. And there was no real analytics. It was all just report writing. And so, I saw that there was a real need for more ad hoc analytics. I saw a demo of Sbase, which was an amazing tool at the time and decided I needed to solve this problem. So I used my relationships with that company to help fund and start my first startup called Mindshare, and raised a Series A of venture capital, and it was off to the races. I sold that company to a company in Northern California. And that's how I got into display advertising. So into advertising, I should say. So one of the first email marketing companies called Digital Impact, bought my analytics startup and we wedded my analytics platform in with their email marketing platform. And that's where sort of, for me, marketing analytics came to play. And we did really great things with marrying analytics, with actually, digital advertising. When digital advertising, again, was very new. I decided I wanted to scale up from just email into, what was a nascent market called, display advertising, which is those little banner ads that you see everywhere across the internet. And the key here was applying analytics to target those ads. So, you're more likely to click on those ads. And so I joined a startup where I used my analytics expertise for display advertising called BlueLithium. And we did an amazing job for our customers such that Yahoo saw us and then Yahoo decided to buy Digital Impact. And that's how I got my entree into Yahoo. So having been a startup guy, both starting startups, working in startups, and then going to, you know, Yahoo, a completely different experience and having to operate in a larger company. But love the scale. So, you know, we went from, you know, three and a half million emails a day at Digital Impact to 350 digital ads a day at BlueLithium to three and a half billion ads a day at Yahoo. So each orders of magnitude sort of increase in scale. And it really taught me a lot about what big data was before that was a term, and how to use analytics to actually really drive in that terms: real operational analytics, behavioural targeting analytics for a market that was still new. So that's where a lot of my experience at struggling with making analytics really consumable by the business - that's where I really learned and suffered, those paper cuts, having to eventually running analytics for Yahoo. So coming in as a small, you know, subsidiary to being able to apply my analytics to a larger business, to being able to run all analytics for Yahoo. And that's where I learned that there was a real gap in the market, which led me to try to solve that problem at a new company called Klout, where we did your Klout score on social media. So I learned a lot about taking digital advertising analytics, and making it work for social media, which was very new at the time. And then that's where I decided ''Hey, I need to go solve this problem for every industry''. It's where I started AtScale with my co-founders from Yahoo, and from some other companies that I had worked with over the years. And so that's what takes us up to current. So 2013, I founded AtScale, and eight years later, here, we have a really great business,
Jonas Christensen 8:44
What a story and what a background. We're going to get to AtScale. But I would love to dig into some of the things you mentioned throughout there. So as a young man, when you're in this startup world, you could call it, and all of a sudden, this company gets sold for so much money - 120 million you talked about. Think those years, the sort of really formative years, and they put subliminal ideas into your head in terms of what can be and the situation to some extent, what do you think that experience did for you, in terms of your further career? I'm almost suggesting that it's kind of a sliding doors moment without recognising it at the time.
David Mariani 9:23
That's a good question. So I sold my first company, Mindshare, for 35 million, and then BlueLithium was worth 300 million. So there was definitely orders of magnitudes that are on the scale in terms of selling the companies. But what I learned a lot is that, you know, when companies acquire other companies, my experience has always been in each of my acquisitions where I'm the acquired entity: The acquirer really doesn't know what to do with you. And they have an idea and a concept. But really, when it comes down to it, it's really up to you to make the most out of it. And so I've always saw that as opportunities to really look at the bigger company and look at the business and say ''What can I do to scale up my talents and my effect, to take it beyond just, you know, my startup and what we could touch?''. Like in the case of Yahoo, all of a sudden, you know, I'm talking to Procter & Gamble and talking about hundreds of millions of dollars of budget. And so it's not just a scale of data in terms of big scale data, but also big scale money and budget. Me also, you know, I wasn't just dealing with my team of 20. Yahoo was dealing with a team of hundreds, scattered across, you know, India, Southern California, northern California and beyond. So dealing with, you know, distributed teams, dealing with lots of people in bigger organisations. So what it taught me is, you know - the startup is all about ''How do I move fast? And how do I break things and fix them fast? How do I learn fast?''. In a company like Yahoo, what I learned is that that scale can be an inhibitor. And so really, it's about building organisations that actually can deliver that agility within the larger company. So the Yahoo experience really taught me how to scale not just on data, but scale on organisation as well. And so it taught me how to organise hundreds of people in teams that could still move very fast. And that was a learning experience. And so I like to attribute that to - you know, I've always been a technology leader, in the case of my own company AtScale: you know, I was the CEO, and I was the CEO for the first five years of that experience. And so I think the Yahoo experience really taught me how to be a CEO, and that I had to do many more things at scale, both organizationally, as well as technically.
Jonas Christensen 11:41
So tell us a bit about that. How did it teach you to be a CEO? What was the stuff that you picked up on there that you hadn't realised beforehand?
David Mariani 11:48
Well, first of all, I had no intention to ever be a CEO. And I was talking to one of my very good friends who worked for Kleiner Perkins. Just joined Kleiner Perkins and very successful in his own right, leading organisations like Twitter and the like. And, you know, I was having lunch with him and telling him about this idea. It's like ''Why didn't somebody solve this problem that I was experiencing at Yahoo?''. And he said ''Well, why don't you solve this problem?''. It's like ''Well, because like, I don't know, anybody who can - I don't know a CEO I can hire''. So, ''Well, why don't you CEO?''. And so just by sort of challenge me on that, it's like ''Well, how am I gonna raise money?''. It's like ''Well, it's easy to raise money. I can help you raise money''. Turned out, he didn't help me raise money. But he got me going down the path of believing that it's something that I can start, and that I can do it. And lo and behold, you know, it's like, all of a sudden, I'm being CEO. I am learning. But you know, having worked in management teams, both small and large, really sort of helped me in two ways. Number one, it's really hard for people coming from big companies to know how to start something from scratch, because you don't have all the infrastructure and support that comes along with working in a large organisation. So I already knew that, having to start something small. But also, people who only work in small organisations also can't see where it's going to go. And so Yahoo allowed me to see where it was going to go and what I needed to do to scale to deliver our products and services at scale. And so having that experience of seeing both sides, how to start and scale and work at something small from scratch. And also then how to scale it out to make it work for larger organisations was really a key element. And so I'm really super happy that by happenstance, I happen to get the opportunity to work in a large organisation and see that firsthand and learn firsthand.
Jonas Christensen 13:38
Yeah, interesting. And the serendipity of these conversations, sometimes. You look back and you go ''Ah, actually, that was just a coffee conversation and that spurred me on to do something big'' or who knows, you might have done it anyway. But that was an important moment anyway.
David Mariani 13:51
I see, Jonas. I had one tidbit for the listeners. It's like - this is somethin that's always been my mantra, which is like ''I'll always have a conversation''. So no matter how innocuous, it seems, and maybe it seems like a waste of time... It's almost never a waste of time to have a conversation. And there's so many instances in my life, where those conversations that when I went into that meeting, thinking was going to be a waste of time or thinking it was going to be something else, later on, you know, it really opened the door. So what I always say is like ''Always take that meeting. Make your time because you never know where it will lead''. And there was so many times in my life like that one, where that was a conversation that was life changing. Where it was just a lunch with a smart guy where we were just having fun.
Jonas Christensen 14:35
And personally, I can subscribe to that. I think the last five years, I've gained a much stronger respect or perhaps the word is appreciation for serendipity, which is what you're trying to foster. The random occurrence, which is not completely random, but things happen when you put yourself out there. So it's really important to do that. For me, this podcast has created some amazing serendipitous opportunities. For example, because you meet interesting people like yourself. So Dave, you've described now your career here, which is really, really interesting to follow in a 20 minute look-back, which has obviously taken 25 years for you to get to this point, or maybe even longer than that. But you describe yourself as starting out as a business guy that then becomes the technology guy, and then later becomes the analytics guy as well. What made you go into the area of analytics and data science?
David Mariani 15:30
Yeah, so I could say that, you know, first of all, the combination of business and understanding technology has really served me, really well over the years. And the reason being is that I'm kind of lucky that I went through the path because being able to understand - if you think about a good product person, a good product person understands the business problem, and also understands the technology possibilities or limitations and how you would actually solve the problem. And so I think that combination of being able to understand the business problem, and then know how you could solve it with technology is a really important combination. And for me, I don't know, data was just always - I love databases. So the relational database was very young. I think I saw Oracle at that first job I had and Oracle was version four, I think. Something like that. And I was just fascinated with databases. Why databases? And that again, I learned dBase, which was a database at the time. And to me, the ability to store information, and then retrieve information through queries was something that excited me, like nothing else. So, Analytics was like baked into me during my first job out of college and knowing how to store and retrieve data was part of my DNA. You know, when I was teaching myself Basic on my Apple II Plus, my first program was a stock program. And the stock program was to track my stocks, because I loved to trade stocks at the time, and I had no way of tracking them. And you had to have a database to be able to store those prices and store those positions. To know how then to report on what their value was at any point in time. So I think I caught the bug very early on.
Jonas Christensen 17:08
So there's a natural tendency or a natural attraction to it there. So Dave, another thing I've noticed, looking at your CV is that it seems you've been able to ride the wave of some really important tech trends that were quite groundbreaking at the time, and including AtScale now that you've started. And could you talk us through those different roles, and what each of them gave you in terms of why that was the right thing to do at the time? The right bet to make, I suppose.
David Mariani 17:37
Yeah. So you know, if you think about sort of the trends, right, so it was really for my path, it was relational databases. So data warehousing, I was very early on in data warehousing. I wasn't in the - ah, I actually was in the mainframe. So our original sort of sources for our financial customers were all Mainframe and COBOL. But you know, I was more on the modern side there, were I was moving that into relational databases like Oracle and Sbase and SQL Server. So I don't know that I have any kind of special skill there, other than seeing transformational technologies. And to me, it's like I say, for you to actually have an idea about what's next, you have to sort of suffer the pain of what's now. You know, coming from COBOL and the Mainframe, so it was very clear that nobody could get access to that data. And a relational database with a query language that could work on a PC was definitely a big move forward, a big step forward. It was gonna allow a lot more people to be able to ask questions. And so that was very clear to me. And that was a clear place that I wanted to be and put my time - was in data warehousing and relational database. Once I got to and tried to scale those up, OLAP technology was really fascinating to me. So seeing what Sbase was doing with being able to do measures and dimensions and drill downs, and how fast it was, and how business-friendly it was. So that next step of then going from rows and columns, were with a very complicated SQL language where you had to be a programmer to access it. And then seeing people using Excel to be able to do and ask very sophisticated questions of data was another sort of monumental improvement in technology. But then I saw that moving to Yahoo, that technology just didn't scale, and it stopped scaling. And so I started seeing people sort of migrate back into the old ways of using SQL and only SQL as their interface to query data. And to me having been seeing the progress of what OLAP did, and how it basically democratised data access, having to move back to SQL because OLAP stop scaling: That was terrible to me. And then that was like ''This cannot stand''. So at Yahoo, we invented a Dup because the relational database stopped scaling. And so that's how we solve that problem. And that was clear to me that distributed computing was the best way to go, rather than trying to build the biggest single computers to run your workloads on. That made no sense to me. So that sort of scale-out technology was clearly the right way to solving that problem. But I wanted to bring OLAP back. Meaning, you know, allow analytics to be accessed by anyone. Not just a developer, not just somebody who can understand the SQL language, but make it work at scale. So it was really sort of suffering that problem, and seeing the trend of what it could have been, and what it used to be. Something that was good that we stopped doing. So how do I take the good that we stopped doing and bring it back and make it work on modern technology at scale? As I was writing the business plan for AtScale, I kept on saying ''At scale. Making analytics work at that scale''. And that's where the name came from. Like ''Why not just call it AtScale?''. But in short, that's a long explanation, Jonas, and answer to your question. But I always say ''You have to walk in the shoes of the customer''. And for my case, I was the customer. I wanted to buy a technology that didn't exist. So we went out and built it. And so to me, it's like, you know, somebody starting out of school, who studied software engineering, and decides they're going to do a startup. I always say ''Good luck. You're gonna need a lot. But go work for somebody else. Go work for somebody else who's solving an interesting problem, and see what problems that they're trying to solve, where they're struggling''. Because I guarantee you, there's an idea there for a new company or new technology. If you're working for a cutting-edge company that is trying to do cutting-edge things, they're going to have problems and they're going to be exposing gaps, where there's a potential market fit for somebody to start and and solve that problem.
Jonas Christensen 21:50
Brilliant insights. So Dave, let's now talk about the present day and your company AtScale. Because I'm sure listeners are dying to hear what is AtScale, what do you do? And what problems do you solve for your customers?
David Mariani 22:03
Yeah, so I'll tell you the reason why I started AtScale. And that's the best way to describing the problems we solve for our customer. So when I was at Yahoo, and running analytics, it was a complex environment. So we had big data, before it was called big data. And so scale was an issue. We invented to a Dup, because Oracle couldn't scale for us. And so that's solved the scale-out problem. But what it didn't solve, it didn't solve the last mile analytics problem. And what we were doing is we're trying to marry tools like Tableau, MicroStrategy, QlikView. And then we had everybody using Excel. We had custom applications for publishers and advertisers. There was a lot of people with their mouths open, waiting to be fed data. And it was my job to feed them. And what I found is that I had to, you know, with hundreds of data engineers, basically take our big data and make it small, and make it small for each of these different analytical applications, these different analytical tools. And what I wanted was to be able to create data as a service once and allow anybody to use MicroStrategy, if that's the tool they wanted to use, or Excel or Tableau, or a QlikView, or an application - build applications with my data. And I didn't want to have to actually create different pipelines for each of these tools. And what I call that as like, is a semantic layer. And that's what AtScale, that's what we ended up building is an independent, universal semantic layer. And what that means is that rather than embedding business logic into the actual tools themselves, like a Tableau, or MicroStrategy, or a Looker or Power BI, is extract that business logic and have it live by itself. And have it live independent of the data it's talking to, as well as the tools that want to talk to it, and talk to the data. So by having that semantic layer, I can get that business-friendly interface available to everybody. And they can use the tool that they want to use to consume that data. So if they're comfortable using Excel, because they're in finance, let them use Excel. If they want to use Tableau, because their marketing use Tableau. But at the end of the day, that Tableau user and that Excel user, they're looking at that data and they're saying ''How many clicks did I have on Yahoo Sports yesterday?''. That number is going to be the same regardless of what tool they used. And by the way, they're not going to have to worry about writing sequel to get that answer. Or knowing where the data is or what the tables are, or know how to write MapReduce code, for god's sakes. They can basically just say ''Show me pageviews by web property, but for yesterday''. And that is what a semantic layer does: is it provides that business-friendly interface to allow people to use the tools they know and love, to access data at any scale. And so we built that for our customers, like our retail customers like Home Depot and Wayfair. Our financial services customers like Visa and Wells Fargo and Fidelity. So our CPG companies or manufacturers like Toyota, General Mills, Tyson Foods. So big enterprises who have big complex environments who want to drive self service and make data available to anyone in the organisation, so that everybody can make a data driven decision. Not just people who are fluent in data, and SQL.
Jonas Christensen 25:24
And I think whether you're a data professional - or may I call it - a regular business person, I think you will be able to subscribe to the challenge that most organisations have with two very fundamental things, which is: Access to information easily, so that you can basically ask better questions quicker of your data. And secondly, just the accuracy and consistency of that same information. So everyone who's listening to this can probably come up with a big handful of examples, where they're spending a lot of time in their day going ''Oh, is this the gross margin? Is that the gross margin? Or did we sell 5 or 10 units of this last month, because we can't agree that there's more than one version of the same''. So this is a really important fundamental problem to solve for many organisations still in 2022.
David Mariani 26:12
Yeah, and I like what you said about information. Data doesn't become information until it has metadata associated with it, that makes it information. And so, I like to think of: Today we've regressed with our new big data tools back to providing people access to data, versus access to information. And so our passion really is to turn data into information, so that more people can use it. Not just a data scientist or an expert business intelligence engineer.
Jonas Christensen 26:43
So good technology doesn't feel like technology. It feels like a great interface, where you're just interacting with what you're trying to do in itself, which is where analytics is still stuck in code in many ways.
David Mariani 26:55
You know, Jonas, though, it's really hard. And the reason why it's really hard -and people told me before I started the company, that that's not possible. And you know, once I started to show the product and the working prototype, you know, the answer was ''I don't believe you''. And the reason being is that for a semantic layer to work, it needs to be universal. Which means you got to be able to talk to the semantic layer and ask questions, using a variety of protocols. Not just SQL, but MDX and DAX, and Python, and REST, and JDBC, and ODBC. So being able to provide all those interfaces and have them work equally well is hard. And then it's also not universal, unless they can connect to any kind of data source. So it's not just on-prem data sources like Oracle, but new cloud data sources like Snowflake and BigQuery. And then data lakes like S3. So it's a big problem, because it's many inputs and many outputs. And you got to be able to cover every combination, which as you start doing the math, it explodes pretty quickly. And so our poor quality engineers who have to test all this have a huge challenge in front of them, because there's a unlimited combination of inputs and outputs that they need to make work and make work at scale.
Jonas Christensen 28:09
Yeah. So you said that people didn't believe it until you did it. And unfortunately, it's very hard to show the tool live on a podcast. But let's try and and visualise it for people nonetheless. So you've mentioned a bunch of clients here that are pretty big and pretty complex. And what I'm imagining sitting behind the scenes there, is a sea of different applications that are from different generations. Old, new, different technologies interacting, and data going across there. And you're coming in to solve this and put this semantic layer on top and say ''Now you all have the same one truth''. What results and benefits have these clients seen? What are some sort of, you know, some good underground examples of what's come about from implementing this in somebody's organisations?
David Mariani 28:56
Yeah, that's a great one. So I'll give you an example. And so, for a semantic layer to be effective - and this is a benefit and a problem - so if done, right, you don't know this semantic layer exists. You don't even know you're using it. It's invisible. And that means that your users don't know they're interacting with a semantic layer. They're just getting business-friendly data using Excel or whatever tool they want to use, or you're a data scientist, and you're in a Jupyter Notebook. It's magic. So it seems like magic. Now, of course, for us, that means that the AtScale name is not out there, right? Because the thousands of users who are taking advantage of AtScale have no idea it's powered by AtScale. Because their interface is Tableau or Excel or that Jupyter Notebook. It's not AtScale. And so let's just talk about one of those big retailers, I mentioned. You think about what happened during COVID and all the challenges with inventory. And what they did with AtScale is a - first of all had AtScale. The first application was to all their store managers. So 3500 store managers, using just an Excel spreadsheet, talking live to AtScale, which talks to BigQuery in the cloud, to be able to get all their sales information across their store, as well as every store in every region across all time. That was not possible before. Before, they could get a slice of data that was static that got updated once a month, and it was only their store data. So there's no way they can compare it to anybody. So all of a sudden, they're using the same tool, using Excel. But now their wealth of data is available to them, that it wasn't available before. So then down during COVID, we have inventory problems. So, you know, basically, the retailer could try to solve those inventory problems by themselves, or they can let their suppliers help them solve that problem. So by exposing that semantic layer to their suppliers, 9000 suppliers, using the same tools: Using Excel, as well as a custom application, they built around AtScale. Those suppliers could make sure that product were in those stores, in each region whenever possible. Because those suppliers, it's in their best interest to make sure their products are on those shelves. So it's in their interest to make sure that inventory arrives at those different locations. And the spread amongst those locations is very different when it comes to inventory and demand. So by having those 9000 suppliers work for them on their behalf, that retailer had no problems with inventory shortages, and they had the right products and the right stores, you know, during the worst time of COVID. Because it wasn't just about them trying to solve that problem. They use their partners and their business partners to help solve that problem for them. And those business partners were all the better for it.
Jonas Christensen 31:35
Great example. And the visual that I have in my head is: I like to compare the maturity of data analytics to the maturity of IT, or the same sort of history or a sense at least. And many listeners can probably remember a time where being computer literate meant that you were able to write DOS commands, because that was how you interface with a computer. And all of a sudden, there was Windows 95 and beyond. And it was very easy. It was dragging and dropping files. And it was sort of a logical interface. And in many ways, the data analytic space is still in a DOS command world. And you're sort of trying to build that Windows 95 interface into data. And you can hear just what it opens up for frontline staff and they can interact with that information without having to become pseudo-analysts before actually being able to interact with that data. Very, very powerful. And so many organisations talk about ''How do we create data literacy and so on in the organisation?''. Well, part of it is making that data available, so that people can play with it and learn from just interacting with it and getting better at it. That's how we all learn with most things. But a lot of the challenges now are focused around this basic accessibility, I suppose. And you're really making me reflect on that in a different way here through this conversation.
David Mariani 32:56
Yeah, you know, Jonas, I call it ''Data for everyone''. That's our goal. It's Data for Everyone. And when I say everyone, it's like, it's everyone. This is like, it's the people who just need to do their job. And then you think about: People forget about Excel. It's like, everybody uses a spreadsheet because it's a model that they can understand. And for years, we've sort of looked, and we looked down on this spreadsheet, but what we did and we said ''Don't use a spreadsheet as a database. it's not for that.''. But allow you to create cells that can point to your data in your data warehouse, in a natural way. And so you can, you know, integrate your own cells of data from your warehouse with your own calculations or your own models. And that is super powerful. Because at the end of the day, whether they're using Tableau or Power BI, they'll always end up exporting it to Excel to do that last bit of massaging, and why do that? Why not just allow them to use the tool that they're naturally comfortable with? So that is what I mean by ''Data for Everyone''. You got to take the data to everyone and their tools that they're used to using rather than expect them to learn new skills to use data.
Jonas Christensen 33:58
Yeah, I'll give you a very simple low level example of that for my career. So I've worked in this space of analytics, data science and data for many years and had different roles there. And at one point in my career, there was a challenge in an organisation where we had too many reports, right. So where do you go to find the information? We had 35 reports just describing one product. And I said ''Folks, we need to clean this up. This is messy. Let's create one, where all this information sits in and you can go and just get it from there''. And - I think she was the head of sales department or something at a time - turns around to me and says ''Yeah, but you know what, I just want that email every morning with that one table in it that says how many sales I had yesterday. Because if I don't get that, I'll forget to go into your tool and look at it. And all of a sudden, I haven't looked at it for a week''.
David Mariani 34:46
That's right.
Jonas Christensen 34:46
And that made me really think about my technical solution. It was not worse - It was great - and probably right to streamline and clean up all that. It was not actually that fit for that user that had a very simple need in one way but their complexity lies in another space, and they just wanted that number every day. So they could track everyday where they were at. Then go and explore further. So again, low level solution: they get a PNG file in an image file, or JPEG or whatever it was in an email every morning saying ''Your Sell sum from yesterday was X''. And that satisfies that need until this problem.
David Mariani 35:22
Yes.
Jonas Christensen 35:22
And that is what we sometimes have to think about. Pull ourselves out of the complexity and solve in simple ways. The way you talk about the marrying up of - I call it principles versus pragmatism. There's a principle of how things should be and how we think the world should work. And then there's the pragmatic view of ''Well, how do people actually want it?'' Right, so people do want it in Excel, so let's not poopoo Excel.
David Mariani 35:44
I love ''Keep It Simple, Stupid'' right? That's a lot of it. The other part of it is like they always say, Don't go into a relationship expecting to change the other person. I always say ''Don't expect to change habits''. If your whole plan is that you're going to change habits of your users, good luck with that. So don't fight them,. Find a way to go to them and talk to them in their language rather than trying to change habits.
Jonas Christensen 36:05
Brilliant.
Hi, there, dear listener, I just want to quickly let you know that I have recently published a book with six other authors, called ''Demystifying AI for the Enterprise. A Playbook for Digital Transformation''. If you'd like to learn more about the book, then head over to www.leadersofanalytics.com/ai. Now back to the show.
Dave, we've talked about these large organisations and the often very complex setup that they have. And various systems, creating and collecting data, data entering at different points in a product journey or client journey, whatever it is we're trying to describe. So these organisations have this very complex environment. How do they set themselves up for success in a data world, both technically and culturally? So could you give some examples of how some of your customers have been able to pull this off and really become data-driven?
David Mariani 37:07
Yeah, you know, my favourite customer, which is another retailer, different retailer. The other example is they treat data as a product. And so what does that mean? So unpack that. What it means is that just like they sell products to their customers, which are very important, obviously. They treat data the same. Which means that they have product managers for data. And they have, you know, engineers that engineer data, and the product managers get requirements. And they create data as a product, which means by creating data as a product, it does a couple things, right. It turns data into a software discipline, which has all the benefits of software discipline when it comes to CI/CD and the like, repeatability, quality and the like. But what it also does: it creates data as a service. As a product that can be consumed by everyone. And by treating data as a product, it makes it a first class citizen on the level of your rest of your products and services that you're in business to sell. And by making data at that same level, it's not a cost centrer anymore. Now it's a core part of the business. And so by doing that, I've seen that that retail company has just many, many more people making data driven decisions, because of that philosophy. And by investing in data at that level. And so I always talk about data as a product. And that philosophy as being a real core goal that I think any chief data officer or CIO should have. It's that by elevating and thinking data as a product, you're gonna spend a lot more time on making it right. And also satisfying your customers, both internal and external. So it's not just your internal users. You got to be thinking about your business partners, because you want to be able to extend your realm, to be able to have them be able to help you make your business better. And so you got to really think of it holistically, not just for yourself,
Jonas Christensen 38:59
Very good insights. And you can see how that - Well, I was gonna call it simple - It's actually not simple at all. That notion of just changing your view on data and treating it as a product. Another word for that, I think, is strategic assets.
David Mariani 39:12
Yes.
Jonas Christensen 39:13
It really then shifts the whole organisation into gear as to how you use that and how you respect it and how you implement it into various processes. And the rest will follow. So almost the assumption then.
David Mariani 39:26
Yeah, well said.
Jonas Christensen 39:27
So Dave, let's talk a little bit about technology here. So we've got these organisations that have their legacy tech and new tech mixed up. But I'm interested from your point of view, because you are very well versed in this space. If you were to design a modern tech stack for an organisation wanting to leverage advanced data science, and you could put in there whatever you wanted, clean slate, What would it look like and why?
David Mariani 39:54
Whoo, I love this question. So if I had what we have today, and I was at Yahoo, this is what I would do: I would build a stack. I got to be careful because these are all are partners. There's really - a semantic layer would be key. Obviously I wanted that at Yahoo and I couldn't find it. So I had to go start a company and build one. And the key for the semantic layer, key principle there is to stand on the shoulders of giants. That's always like to say. So the semantic layer doesn't hold data. It is really an API layer, sits in-between the tools and your data. And what that means is that your stack with that semantic layer is that you can support multiple tools. So like I said earlier, I don't expect to change habits. So I need to be able to support as many inputs as many sort of consumer types as possible, including the data scientists and the business analysts, and the application developer. So I'd make sure I have a semantic layer for that. And then I would choose a cloud based data platform to power that. I would make sure - without mentioning names - that those cloud data platforms could be used on different clouds. So I wouldn't choose cloud proprietary technologies, because then you're wedded to that cloud provider. And I think that's a poor choice. So I would choose tools that could work on multiple clouds, so that when I did have data on multiple clouds or decided that I wanted to change my clouds, I wouldn't have to retool my whole infrastructure. There's a whole argument going on about a data warehouse versus a data lake house. I think that's all hogwash. Of course, you have a data lake. That's where your data lands. And then you have a data warehouse, which is what makes it fast and provides a great interface for analytics. So I would have both and I would allow access to both. Your data lake: People need access to their data lake, like data, scientists need granular data for training their models and doing exploration and exploratory analysis. So you need access to the data lake data, which is raw. You'd also need access to the clean data in the Data Warehouse, because finance can't be dealing with raw data. And so to me, it's not an either/or. It's an and. So, I would have a cloud based data platform infrastructure powered by a data lake and a data warehouse. It would run on any cloud. I'd have a semantic layer, and then would have whatever tools were popular in the day. And I have another set of tools I can't tell you, because they're all partners of ours. I would not try to force my users down a single path. I would support them again, wherever they lay. And that would include Excel, by the way. And not just an Excel as a database, or a data dump, but live connections from Excel, so that anybody with Excel on their desktop could be a consumer of my data.
Jonas Christensen 42:33
And I love this marrying up of existing technologies and existing skills with the modern tech. It really also for me highlights what you're trying to do at scale and and what the product is trying to solve. It's trying to some extent - and I'm putting words in your mouth - but it's trying, to some extent fix this challenge that a lot of organisations have with legacy systems and new systems trying to interface and integrate data across the US. Is that fair to say?
David Mariani 43:02
Yeah, we took an integration-first strategy. And that's always been my philosophy, and even running the analytics organisations integration first. If you think about a lot of our competitors who don't exist today, they built full stacks. So they built their own database with their own analytical engine with their own visualisation tools. And they said ''Hey, Mr. Customer adopt this. So forget about all the other stuff you've invested in. Here's this new stack, that's better''. And it doesn't matter if it's better. It's different. And again, you rely on the customer changing habits. And to me that is the wrong approach. It's that instead, you should take an integration approach. And you should make your technology and your technology stack as fungible as possible. As plug and play as possible. So that you can isolate your dependencies. You can firewall those dependencies, because you never know what's going to happen in the future. Because I can tell you, there'll be a new BI tool or a new AI auto ML platform that someone's going to want to bring into the organisation because ''It's the best''. There's going to be a new way of storing data because ''it's the best'' and you're not going to be able to fight that. Maybe you don't even want to fight that. So you've got to have technology that will allow you to be flexible to be able to adopt new technologies without retooling the entire stack. And so that's really the principle that I would sort of hold when I look at developing a resilient analytic stack for a modern organisation.
Jonas Christensen 44:31
Brilliant. Thank you so much. Now, Dave, we're coming towards the end of the conversation. So I have saved that one big question for you here towards the end. What's your vision for how data driven organisations will function in the future? And how does the universal semantic layer fit into this picture?
David Mariani 44:49
Yeah. So I think the universal semantic layer is really the unifying thread that joins what today is - we have really two different silos of analytics consumers. I would say three. You have your business analysts. You have your data scientists and then you have your application developers. And they're all in their own little world using their own tools and stacks. And they all exist under the same roof in the same organisations. So if you think about the data scientist, the data scientist needs to consume data just like the business analyst does. But they also are predicting the future. They're generating new features, new predictions that that business analyst needs, and that the application developer needs because they need to operationalize it. So they can order that new product or new inventory, when the model says to order it. And so to me, the semantic layer is that unifying thread that brings these teams together, so that the business analysts can not just look at what happened yesterday but also be able to look at what's going to happen tomorrow, because the data scientist has predicted what's going to happen tomorrow. And those data elements are consumable in that semantic layer. And that application developer can now operationalize those decisions, to do prescriptive analysis to actually take action automatically, autonomously on that data. So my vision of the future is that analytics is no longer just a human-led endeavour. But we're able to bake analytics powered by humans, but into the operations of an enterprise. And that decisions can be made with confidence that take away some of the drudgery that we currently put on the shoulders of our business users and our users in the organisation, where much more of our decisioning can be made autonomously, versus finding the needle in the haystack, and then figuring out what to do about it.
Jonas Christensen 46:39
Brilliant. And you're making me reflect on the last 10 years of analytics maturity across many industries. And I think, whereas 10 years ago, there were challenges in terms of speed and scale of big data, our ability to be accurate enough for predictions, and so on. Those things have largely been - not solved, but - they've been improved substantially with cloud technologies. And the further advancement of MLM and so on. The biggest challenge now is actually operationalizing some of the stuff and getting into the front line. So we can all build fantastic models, but actually getting it into a business operation that is firstly meaningful, but then accurate, and timely is the biggest challenge. And that's what your semanticics - I can really see it for myself how it can help solve that challenge. Now, Dave, before we finish up, I've got a couple of questions for you. And one is one that I always ask of guests on the show and that is to pay it forward. So who would you like to see as the next guest and Leaders of Analytics? And why?
David Mariani 47:39
Yeah, you know, I've been following Benn Stancil. He's the founder and chief analytics officer at Mode. And he's been writing some really interesting blog posts. He has a great blog series. Talking about the challenges of data and talking a lot about, you know, how to make data more approachable. So I think I'm older. So I have more history and also biases sort of baked into me. I think he's more born in the modern age but also thinking about the same problems of ''How do you deliver semantics?'' or what he calls metric layers or metric stores. So I think he's an interesting person to talk to. I would love to listen to a conversation that you might have with him.
Jonas Christensen 48:16
Brilliant. So I will definitely be reaching out to Benn. So thank you for that recommendation. And lastly, Dave, where can people find out more about you and get a hold of your content?
David Mariani 48:27
Well, he's it's very easy: Atscale.com. We have great stuff. When it comes to white papers. We have use cases and case studies with real customers. We have a webinar series. It's a thought leadership series that we run once a month, where we don't talk about AtScale, but we talk about the challenges and we have real analytics leaders talk about how they solve those challenges. There's a really great amount of information about our industry and what's happening next. And if you go to the Atscale.com Resources page, you're gonna find it all right there.
Jonas Christensen 49:02
Yeah. And in my research for this podcast, I did look at some of the stuff you have on there. There's a lot of interesting stuff and a lot of golden there. So I do recommend listeners to have a look at the website. And as you can hear, Dave is just a passionate nerd in this area. So you're just gonna get real value out of it. It's great, what you're doing. I think in terms of creating a future world for us all here that is much more data driven. David Mariani, thank you so much for being a Leaders of Analytics today. It's been a true pleasure and a lot of value has been given to listeners, and I've learned a lot personally. So thank you so much, and all the best for the future.
David Mariani 49:39
Thanks, Jonas. Thanks for doing this and thanks for having me on.
