Transcript
Dave Mariani: Hi everyone. And welcome to AtScale’s data-driven podcast. And today I have a very special guest. I have, Benn Stancil and Benn is the chief analytics officer and founder, for mode. So, Benn, welcome to the podcast.
Benn Stancil: Thanks for having me. It’s good to be here.
Dave Mariani: So, so then there’s, you’re a, co-founder, you’re a founder like, like me and, you know, I always love to hear, about the, the origin story of, of mode. And if you could just tell, tell the audience a little bit in the listeners about how you got to where you are and, and, and really how mode came to be.
Benn Stancil: Sure. Yeah. So a little bit about myself. So I, afterschool, I actually worked completely outside of tech. I worked at a think tank in Washington, DC, doing very DC stuff of like writing policy recommendations that nobody read. I ended up, ended up in a job in San Francisco a few years after that. looking basically when it, out of that world, but I liked thinking about problems the way we thought about. So our job was to basically look at, look at like the macroeconomic situation in the world, and do a bunch of analysis on things and make recommendations. And, and to me, I was a million steps removed from anything actually making a difference like Ben Bernanki did not care what I thought, even though I was writing things to tell Ben Bernanki what to think. and so I wanted to keep thinking about problems that way, but not in a way where I was essentially just yelling into the void. and so I ended up getting a job in San Francisco, through like a friend of a friend, essentially knew somebody who knew somebody who put me in touch kind of stuff, the way this stuff happens, as a, as an analyst on a data team. and so it was company called Gammer. it was an early SAS product for basically communication at work. It was sort of similar to slack or Microsoft teams.
Dave Mariani: Yeah. I remember a Yammer Yammer.
Benn Stancil: so yeah, so it was, it was like relatively popular. It got acquired by Microsoft in 2012. and so after that had happened, I, on that team, we had built a bunch of internal data tools that our job was what is now become kind of a common thing for a lot of analytics works, which is go help other people in the business make decisions. and we weren’t a statistics team. We weren’t like a bunch of a bunch of PhDs and in some sort of quantitative fields sitting around trying to do really complicated bath. And we also want like a BI team that was just building reports. Our job was somebody in marketing needs to figure out where do we run our next campaigns or somebody on product and decide what to build next, go apply a bunch of data to it and help them think it through.
Benn Stancil: and so, so that’s now how I know is the inner shape, but it wasn’t really what they looked like so much at the time. And so we didn’t have any tools to help us do that. Like we didn’t want the sasses or the status or like the really technical stuff, cause that wasn’t really what we needed, but we also didn’t want to use just like the MicroStrategy business objects, BI stuff, because our job wasn’t to build dashboards, it was to help people make decisions. So we ended up building a bunch of internal tools ourselves kind of core among them was essentially a bright sequel in a browser, make a chart, send a link to somebody. They can see the chart. It was like a way for us to collaborate and facilitate conversation for other people in doing that. Me and a couple of other folks realized that a bunch of other people around Silicon valley were actually building the same thing that Facebook had a version of this Airbnb and Pinterest and LinkedIn and Spotify, and all these companies had built these like internal query tools designed to help their data to do exactly.
Benn Stancil: and so that really, once we sort of saw that we’re like, wait a minute, this is actually a thing that there’s closing demand for. There was also a lot of the early stages of things like Redshift were starting to pop up where it used to not be accessible of like being able to do analytics on, on data warehouses, which was primary. You cost a hundred thousand dollars a year to support things like Teradata now is very accessible. And so given those two things, we were kind of like, Hey, if, if this is the thing that is accessible to a lot of people, it’s a trend that leading companies that think about data are starting to push. There must be a market and there will be a market for data teams that are trying to think of problems in this sort of way. and so that’s what we set out to build was like, how do we build a product for analysts and data scientists to help them distribute their work on the organization over time That that core vision has always been more or less the same is evolved to be much more expansive where we realized actually it’s not just the data teams that do that work. It’s a much wider set of people who are thinking this way and the whole entire business needs to be involved in this, not just helping a select view. but that is, that is ultimately what, what mode was and where it came from.
Dave Mariani: Yeah, it’s a, it’s a common theme on this podcast. I talked to a lot of analytics leaders like yourselves and some founders and it’s, mostly analytics leaders don’t come out of, come from a technical background. I mean, they don’t come out with CS degrees in the like it’s like they were, they seemed to come from the business of having to sort of solve the problem and then figured out technically how to do so. and then ended up leading a, a more technical team and doing analytics. So it sounds like you followed a similar path there.
Benn Stancil: Yeah. And that’s, I think we there’s this notion in the data community about like purple people, which I intentionally and very much, dressed in the theme of, go Hornets. They’re planning their one game playoff game today. they, this idea of like, these are people who sort of straddle the line, that they need to be the bridge between the purely technical folks. And I’m not sort of in terms of like an engineering technical, but just like how data gets generated, the technical definitions of where this stuff come from and the business users who are thinking about the actual problem that matters. but don’t actually have that much access into where did this data from, how do we interpret it, that kind of stuff. And so, so our data team at Yammer very much sat on the middle of that, where it was our job to understand that this is where data comes from, but also how do we translate that into a business problem and so I think that’s very much kind of the space that node sits in and then very much where these analytics teams in general now actually operate as, as these kinds of translators between the technical side of the business and, the, the kind of like actual business problem that needs to be addressed.
Dave Mariani: Yeah. You know, I want to talk a little bit about the purple people, cause that’s a, that’s a really interesting topic on its own, but before we, before we go there, I want to just drill down a little bit more about yourself. So how do you find the time by the way So, so, so listeners, Benn has a, a weekly sub stack, blog that is, it’s not just a stream of thought blog. you got some real data in there, all around data and analytics. It comes out every Friday. and I, I read it religiously because there’s always, there’s always something interesting to learn from your writing. How do you, how do you end up, first of all, finding, you know, figuring out what the topic is going to be, and then finding the time to put and to deliver a product like that every week and do your day job.
Benn Stancil: so on the first, the first question there is like a tip. The second question, there’s not really the first question of how do you come up with topics and things like that. There is a, there is an element of sort of the rich get Richard on this, where if you start to do a little bit of this, people talk to you and they’ll ask you questions and you’ll have conversations like this. And it’s really easy to start once you’re once, you know, Hey, I’ve got a thing, I’ve got it. Right. you’re paying a lot more attention to like what had happening. And these are the conversations. Are you sort of catch onto a few things like, oh, this is kind of interesting idea that connects to the other thing. It’s like, it’s always a little bit in the back of my mind. And I think because one that makes you a little bit more sensitive to seeing how the dots get connected, just not specifically, just like, as you’re thinking about it, it’s something you notice.
Benn Stancil: the topics are a little bit easier to come up with that way. And also being able to be in these conversations and having people ask you questions or send you emails or whatever it is is something that means I don’t have to power the whole thing. you know, other people are sort of nudging and directions. I can, I can write it down, but, but the original idea doesn’t have to come from something that I sat in the closet like this one, and came up with, on, on the, like, how do you actually get it done The kind of honest answers is just kind of a lot of work. like there isn’t, there are things that I’ve figured out about writing process that works for me. Actually, I had a post about this kind of back around the new year. That was like, here’s how I do it.
Benn Stancil: Figure out things that work for you and things that adult. but there isn’t really that much of a tip there. I think that, that this Friday bit has become a little bit of an accidental thing that that was just kind of the, the groove of fell into it has been helpful because it provides a deadline. and hitting that deadline. Sometimes it’s just like, oh, this is going to be kind of hard. there’s not like out here as a secret thing to do is to make you do it. It’s like, there’s one tip part of it’s just like, it’s a bit of a grind. but I don’t know, just, it’s interesting. It leads as mentioned in conversation. Sometimes they’re not hard. Sometimes they kind of are easy to you’re sitting down and you’re doing it the right time. The one thing I would say that I have learned in this that is like, maybe the tip is some days you have it, some days you don’t, don’t try to force it when you don’t have it, that you can sit there for four hours and just stare at one paragraph and get nowhere and walk away from it.
Benn Stancil: Watch a bunch of tech talks come back in an hour and you may sit down and write the whole thing in 20 minutes. and so that doesn’t mean it’s always gonna be easy, but it means like if it doesn’t work, just like walk away from it, it’s actually better to get away from it. They try to like
Dave Mariani: Really
Benn Stancil: Slog through it. Cause sometimes you just get stuck.
Dave Mariani: Yeah, I can, I can definitely appreciate that. same thing, you know, but I can tell you though, it sounds like you, when you created the flywheel, right So by putting your content out there, you’re getting responses and you’re creating a community, that community is coming back and giving you more ideas and more content that you can talk about given the kinds of questions, the kind of conversations that you’re having online. that’s, that’s pretty powerful. cause you, you you’ve created it whether you know it or not, then you created a great community. And what I’m saying to the listeners is like, you should definitely, definitely subscribe to Benn’s, blog because it’s, it’s, it’s, it’s, it’s, it’s worth the read every week. never disappoints. So, so you’re, you, you obviously were able to get, get those kinds of insights from that community you created and I’ve, and I’ve been part of that in terms of asking questions, making comments, which is, which is great.
Benn Stancil: Well, I appreciate that. And by all means, continue to ask, as I said, that’s, that’s the, that’s ultimately the, the mind from which the ideas come from. So I appreciate that.
Dave Mariani: Okay. Well, so, okay. Let’s get onto some existential questions. you know, we had a conversation, just a while ago and you probably see now you kind of brought up the, you know, what’s the point of all this, you know, is, is, is, is all that we’re doing analytics isn’t really worth all the effort. So what did you really, what what’d you mean by about that Let’s talk a little bit of a dive deep on that.
Benn Stancil: So yeah, so there’s like, if you ever see like a fundraising pitch from a data company, there will inevitably be some bit in there about how data-driven companies win, how the companies in the modern economy need you to think about. It’s good to like that that notion is kind of embedded in all of this and it’s never really questioned. It’s never really, it’s going to take it out of a matter of faith that like, yes, we are a better companies because we do all these sorts of things. I think that’s probably true, but it’s not as like, obviously true is I think we probably act like it is. and, and it’s not, it’s not like data is universally. Okay. I think it’s just more, it’s useful in, in particular ways, not so useful in other ways. and so I think we have to just be a little bit more or more have to be, but like, and thinking about the existential question of all of this, it’s like where actually has this made a really significant difference and where has it not
Benn Stancil: And I think, I think it has made a really significant difference in operationally in terms of how tight businesses can run and about how well they can respond to things and how well they can understand the world around them. I don’t know that it’s quite yet made the difference in terms of like us making better decisions because of that. I think at some places, yes. But the idea of our company’s like way smarter today than they used to be. Maybe, I don’t know, like I never sort of thought about this before, but there’s, there’s this like a bit of like, couldn’t LeBron be bill Russell, how would they compete against each other you know, how would today’s apple compete against GE from a hundred years ago Like with apple just crushed them in terms of how smart they are. Like maybe, but I don’t actually know that, like, I’m not sure that I entirely buy that. So, you know, I, there’s a question of like, is, is better. Decision-making actually competitive advantage here. And I like maybe I think operationally data helps a lot. It’s better. Decision-making maybe fuzzier.
Dave Mariani: Totally. I totally agree with you on that. Like, so, you know, there’s the old story about, you know, Rick Barry with the sort of granny style free throw shooting, right. With underhanded free throw shooting. And, it’s proven that that was more than 90%, you know, I think he had like a 96% success rate when it comes to like underhanded and, and, and others sort of did it and gotten the same kind of, improvement going from, a very poor free throw shooter to a really good free throw shooter, but they abandoned it just to just the same, just because it’s embarrassing to do brand new style. So it’s like, it doesn’t mean it doesn’t matter, but sometimes the data says what you should do, but people don’t necessarily follow the data when it comes to behavior. but you know what, you know, Benn, you kind of bring up a good point is that it’s, it’s really a sort of fuzzy, connecting sort of outcomes to the, to the investment in data and analytics.
Dave Mariani: And, you know, one of the things that, that the reason why I was so attracted to, to digital advertising was that it very much was tied to data. and you know, and because of behavioral targeting, because you can measure the performance of one ad format in one site versus, or, or in one placement. and, and you can look at the targeting, you can do AB testing. It’s the data, it’s much apparent what works and what doesn’t work. and for the team at Yahoo, you know, they were able to quantify it in terms of real dollars and the dollars made sense. But for, I think that for most a business today, those sort of, that direct sort of correlation just isn’t there. I mean, or it’s very difficult to calculate. So you just, so people are sort of going on pure faith that this investment in analytics is going to be worth the effort.
Benn Stancil: Yeah. And there’s, there’s another part of it too, that I think makes that even harder, which is on, on data and advertising performance. It is a daily, we marched forward, right Like we, we do this, these ads are a little bit better, or we see the results. We see the money that comes in. if you are in sales, you can go to work. You can leave at the end of the day and say like, I can see the progress I made. I advanced these deals forward by a few inches. I like this one got close to the finish line. Like every day is contributing something. And like you mentioned, sort of the blog, like that is something that is a slow build, right You write a blog post, you host a podcast. Every episode that you have kind of builds the thing that you were trying to create.
Benn Stancil: And there is a sense of this gradual progress with the work that folks do on data team. It’s not like you could work for six months and do nothing. And then one day do something that’s very valuable. And so it’s just like very spiky thing that makes it way harder to actually feel the ROI. Like one of the lowest ROI things I’ve ever done in my entire life, knock on wood is wear a seatbelt. I’ve never been in a car crash of any significance. I’ve never had to wear a seat belt in a way where the seatbelt made any difference to my life. And like, every time you get the car, you put it on, but that doesn’t mean I’m gonna stop doing it. And I think like there is some element of that and, and how like data works where it’s like, the reason we do this is in part, it’s not necessarily prevent disaster, but it’s because the, the times when it matters, it can matter tremendously.
Benn Stancil: And so it’s hard to like, what is the ROI would be wearing a seatbelt I don’t know. I guess it’s zero, but like, is it really what I do at differently No. What, and so, you know, I think that’s, that’s, there’s a lot of that and how we make decisions with data where sometimes it’s you make one decision one time and it’s like, this changes the entire trajectory of the business and it made everything else work. And you don’t know where that is. Like, you know, you’re 50 of the things you’ll do will be like, whatever. They don’t matter that much, but you don’t know that until you do. So I think it’s like, it’s also structurally just a very hard thing to measure in that way.
Dave Mariani: Yeah. You know, I’ve used, and this is a very course measurement. It’s really a measurement of demand more than anything else. But, you know, if you just, if you create a data service and you create data assets or data products, and it’s about usage, if people are using them, then you have to make the assumption that they’re getting value out of that use. and you should see an up into the right. And when it comes to usage, if you, if you’re being successful. So I don’t know, you know, like it would be great to have a, I know I ask people this all the time about how they calculate their ROI for their investment in analytics. And it’s always a hand wavy kind of an answer. and the only thing I’ve really heard that makes sense is are people querying your data assets more than they were yesterday I don’t know what else to do. Yeah.
Benn Stancil: And I’m, I’m, I’m with you on that. And like that, that seems less of a problematic answer than a lot of people will be like, well, but is it useful It’s like, well, who cares
Dave Mariani: Like I don’t,
Benn Stancil: Yeah. McDonald’s, doesn’t ask if, like, what was the, was the hamburger I sold you valuable Like, did it really make a difference It’s like, no, you bought it. You must’ve liked it. That’s our job.
Dave Mariani: I like that. That’s a great, yeah.
Benn Stancil: We have to have some faith that the people are, are spending their time wisely, that they are buying the things that they want. And like, if a bunch of people in a company are choosing to buy data assets, I think we have to have some faith that like, these people are actually doing the right thing. and they are, we have to trust their word. Like half of us are economists w you know, trust their rationale, but they are rational actors. They wouldn’t use it.
Dave Mariani: I’m an economist, I’m an economics UCLA. So, yeah, you gotta, you gotta trust the market. Right. So I think that’s good. okay. Let’s shift gears a little bit because I, I hear a lot of from our customers and the customers that I think are the most advanced and the smartest of our customers. They’re, they’re really looking then at this more of a hub and spoke style of, of, of creating data products. So they have analytics. and there’s been a lot of talk about data mash, which I hate that word, because it sounds like Federation, and it’s not, it’s an organizational, it’s an organizing principle. It’s not a technology. but, but it’s the whole, like the whole concept of look we’ll create data products. We’ll, we’ll, we’ll allow the business, to have the have domain ownership of certain data products. So shipping owns shipping because shipping understands shipping, merchandising owns merchandising because they own it. They understand merchandising, finance, owns finance, et cetera. what do you think about that concept of a more decentralized way of, of creating data products in these enterprises Do you think that’s going to be successful Is that, is that something real or is it something that just a, a bunch of analytics and, intellectuals are just chatting about
Benn Stancil: There’s a few ways to think about that. My view is at a big enterprise. It’s a very uninteresting statement because basically what that’s saying is like one giant centralized team can’t serve the big enterprise. It’s like, of course they can, like, of course you have to have some, you know, if you work at Amazon, of course, AWS is not going to have the same, like analytics infrastructure as whole foods. that’s not an interesting statement. The interest excitement is like, how far down do you go And I think it really did companies again, it’s like, there’s a certain scale, or just doesn’t centralization doesn’t make any sense. Like you can’t make it work. So, so the question for me is like, should a 500 person company follow that Or should they have a centralized thing
Benn Stancil: I think the answer to me is I am, I, I think a believer in the human data mesh, but not the technical one. There is, there is some of like the data mesh version of this, like the way people sort of the paper about it, which is a little bit esoteric in, hard to follow is, is it’s a technology, right Like, it’s like, there’s somebody who is building a data asset and is shipping this piece of technology to other teams. And they are the ones who are responsible for figuring stuff out. And it’s, it’s like, there is this, this mash of technology underneath it that’s like services and all that stuff. Okay, fine. Yeah. That seemed a little bit far-fetched to me the idea that you, that you would decentralize once you get to a big enough scale, such that there are people who are primarily thinking about those problems and those people are further up the stack than you would expect kind of makes sense.
Benn Stancil: So like, do you embed analysts into say, are fibers per person coming to do a better than analyst in the marketing team Yeah, probably. I think that some people don’t, but I think like, certainly there’s reasonable case with him in that. Then the question is like, do you embed like an analytics engineer or a data engineer in the marketing team And I actually think that makes some sense that there are problems with how you would actually want to like gather and collect and model and all that stuff. The data that is in the marketing world versus in a product world, especially having someone who dedicated jobs to think about the marketing side of that, as opposed to the product side starts to make some sense. And then you might have someone who’s kind of oversee, you know, somebody who kind of helps coordinate, but ultimately like, is that a data mesh is that this model where you have somebody who’s like in sort of specific terms, like there’s a marketing DVT and a product DVT, and those things are managed by different people. Like, I think that’s a data mesh, but it doesn’t feel that revolutionary to me, it just feels like the same debate we always have about centralized. Should a team be, except the question now incorporates like analytics engineers in addition to analysts.
Dave Mariani: Yeah. I’m thinking Benn, you know what, I, I like the term hub and spoke better because you do have to have a hub and the hub is the hub is going to be responsible for choosing the tools for creating the standards, creating the, you know, creating the rules of the road. Right. and even some of the common assets. Right. So, because look, everybody needs to deal with a date dimension cause so everybody’s talking to the same thing about the fiscal calendar versus the, just the, the, the regular calendar. And, people are gonna have a common product catalogs that they’re going to be using. So there is some need for some team to basically set the rules of the road, choose this tools and the stack choose the processes. And also potentially even some of the core data assets, but the business definitely than domain owners know their business.
Dave Mariani: And no one from the central team is going to be able to learn the whole entire business all at once, especially for complex businesses like we have today. So I do think that the hub and spoke makes sense. I’m not so sure. I, I don’t like the concept. I don’t like the words data mesh. I think it’s very, I think it’s very, it’s, it’s very confusing because it sounds like a technology and it’s really, it’s really more of an organizing principle that I think is bad would be valuable. And so, so anyway, so, so w what do you think about that Do you think there needs to be a central team to, to sort of manage standards, pick tools and the like, is that work
Benn Stancil: Yeah, I, so, so I guess I would put there, there are call it four piers of things. And the question is, where do you, where do you put the centralization versus decentralization line The first tier is the bottom of it’s like, who’s choosing, we use big query versus Netflix, who like that needs to be centralized. Like you can’t have like the marketing team, who’s one database and somebody else do some, right. Okay. You’ve got consistency and tools, consistencies and standards. I don’t think anybody would really argue that unless you get, I mean, maybe there’s some, like everybody can choose anything, but that seems kind of wild. The, the tier up from that as sort of the analytics engineer model, the business part of it, where, okay. You have the data, and now you’re trying to actually like, sort of construct some kind of business logic around it. The tier above that to me is like the analyst themselves who are writing queries and building stuff. And then there’s like the business consumer tier of the, the marketing analyst, the person, not, not some of the data team, but someone who’s like demand gen analyst who just happens to have some of my quantitative abilities.
Benn Stancil: I think the bottom of that clearly centralized top of that, clearly de-centralized like, that’s, that’s a person on the marketing team. that’s the person that makes the decision, right The question is, what about those other two Like, and I think we’ve long had the debate of do analysts centralized, decentralized like that as a very tired debate to me, all the data mesh does like that whole concept is introducing. What about this analytics engineer piece Does that also get decentralized or stay with the centralized team And I think that that is the entire conversation to me is basically that where sure, you could, you could say the analysts themselves should be centralized in which case, obviously, none of this makes sense if that’s your stance, like none of the data mesh stuff is patters. So basically you’re saying, yeah, okay. We should have these centralized analysts. We’d have a centralized technology. What about the analytics engineers Where do they sit
Dave Mariani: I like that. I like that model that you’re thinking, and it’s like, look, it’s pick the two extremes. It’s the answer is obvious there it’s like, and then you fill in what’s in the metal. yeah. W we have, one customer in particular that, is actually abolished the, the whole title of, of analyst, because they expect everybody in the business to be able to be an analyst to do so to be that, that, that person that you said on the that’s embedded in the business, that they need to use data products to make their decisions. And so they’re still as a data engineer in the, in the, in, in the business, but there is no data analyst who’s, who’s helping out the business. It’s the business person is it needs to be able to use the, the analytics that the data engineer produces, as a data product. and that’s still concept
Benn Stancil: And yeah, and I, there’s a, I feel like there’s a, this is, this is a thing that is, there’s a trend in this direction of the analysts getting squeezed basically where there is pressure from the, from below about analytics engineers, where it’s like, oh, analysts actually should just be writing transformation all the time. and then there’s pressure from above around things like self-serve and how do we enable these folks and like data building data products, and, you know, the data team has a product team type of ideas. There’s an easy way to see that where, like the analyst gets squeezed out where it’s like, oh, we should enable everybody to be so literate in data that they can do it themselves. And then they’ll kind of transformation, semantic modeling stuff happens with an analytics engineer. And now we have nobody who’s technically like whose job it is, is to think about the problem. I kinda think that’s a tricky spot. I don’t know that that makes sense to me. like there’s a lot that goes into being good at thinking about the problem that is enabled by tools. so, you know, I think, I think there are some cases where that’s not true. I think some cases where it’s like that would actually simplify things to get rid of that I’ve used this analogy with like Yelp, where we don’t need an analyst, help us decide we’re going to restaurant Yelp gives us a good enough tool to figure it out, we’re to figure it out. Right. But that’s not the problems businesses are solving. They’re harder. And I think you need some places where there are people who specialize in this kind of thing.
Dave Mariani: So Benn, you, you said something interesting though. Cause you said analyst engineer, like non data engineer. So, and there’s, we don’t talk about BI engineers anymore. So
Benn Stancil: What,
Dave Mariani: So when it comes to titles, is that what we’re seeing is we’re seeing really data engineer is really an M and a, analytical engineer. What do you where’s that cause, cause that’s part of the that’s part of the convergence, isn’t it And in that sort of analytics delivery stack, that compression that we’re seeing.
Benn Stancil: Yeah. So, so to me, an animal like DVT basically is the, the company that made these things different title. both in terms of, in terms of the product they built is sort of what enables them, but also they’re, they, they sort of tried to own that, that space and they have to this point done that my, my take of what that really it’s a BI engineer, like it is a BI part, like the, the job in tooling that is better than that and, and sort of less sort of depressing to work in, but ultimately like the thing that an analytics engineer does, as I understand it is you have to work with the data that already you already have in place. A good data engineer is great. We need to put data like we are selling a bunch of products and we didn’t have no idea how to actually keep track of it. We have to like do the work to figure out how to get this data in a place where anybody can make sense of it. the analytics and is basically like the person to, to imbue that data with some kind of like business meaning,
Dave Mariani: meaning.
Benn Stancil: So yeah, it’s like we have a bunch of raw logs. How do we turn this into something maybe anybody can make sense of And like, obviously there’s a lot of fuzziness there between what an analyst does cause an analyst would presumably kind of do the same thing with an analytics engineer. Like I think the argument is you stopped earlier. Like you basically say we have imbued it with logic. Somebody else figured out what we do next, like make the decision with it. That is not my job, which I think could make sense, but I don’t think it’s always necessarily straightforward to say here is the cleaned up data with all of the logic applied to it. Therefore anybody can just make decisions on it. Like there is a process of actually interpreting it. That is hard. And that’s, that’s ultimately like what an analyst brings. I think that is necessary. In some cases it’s not necessary if you’re trying to choose a restaurant on Yelp, it’s probably necessary for trying to choose, like how do we reprice a product
Dave Mariani: So, so you just, you just, you just brought trans transitioned us into the, into the topic about, you mentioned DBT, DBT just did a big funding announcement. They announced that they’re, that they’re delivering, working on delivering a semantic layer. And when you think about that, that analytical engineer, you know, in, you know, imbibing the data really it’s, it is creating business meeting on top of that data. Isn’t it So is that, does, what, what’s your opinion on semantic layer metrics, metric layers, some people call them metrics, stores. What’s your idea about how that, that, that concept and where that fits really in the organization as we took this has been talking about it.
Benn Stancil: so I’m obligated to say that I’m a fan of, of metrics layers. but I think the specifics of a, both of them that make sense, actually. So there is I think the need for a semantic layer or something that is at a semantic, semantic re rendering of the data from something that is sort of a machine generated mass to how do we, as humans understand yes. I think that is true. I think that is one of the, one of the four things that data team or data technologies actually do. Like basically to me, every data technology section may be a thing it’s talking about this week that the technologies are all like fundamentally one of four things. One of them is governance. It is cement. It is all right. We have a thing. What do we make Like, how do we make sure everybody leads this the same way
Benn Stancil: And in that I think like what DBT is doing makes sense what other folks are doing. They’re like there needs to be that. And the, the, the data technology is complicated enough now such that having something to kind of can be universal so that all of the things that read from it can make the sense of the same things in the same way as it is a very valuable problem. And someone needs to solve how specifically you get there. I think that’s a different question and there’s a lot of different potential directions you could go and like ways you can actually implement this. I don’t think that we have had that conversation yet though. Again, this may be a thing for, for this week. there are people are moving in those directions, but I don’t know that that’s, I don’t know that part’s been figured out yet. people have different, but
Dave Mariani: Yeah, definitely look it’s, you know, of course I’m a big believer in the semantic layer, goes to stigma earlier on it. And, and I think it does go back to, that the, our previous conversation about data mesh look, it’s like, if you want the business to be analysts, then you need to speak business to them. and, and it’s almost like, you know, whenever they asked me about the future of AI, to me, the future of AI is like, you don’t know you’re using AI. there’s no such thing as a auto ML machine learning and clusters. And it’s like, you know, the success of AI and ML is when, you don’t know you’re using it. And I, and I think it’s the same thing with data and analytics is that it’s just, it’s just the job you do. It’s part of your job.
Dave Mariani: And you’re not a business analyst, you’re just a business person doing their job. and I don’t know how you do that. If the interface to data is a technical one, I love your Yelp, your, your, your, your, analogy. It’s. So it’s a good, a good example of, Hey, you know, yeah. We can choose the best restaurant given the data presented because it’s an interface that works for making that kind of decision. not all businesses decisions are that simple, but, it’s, it’s, it’s a direction I think that we at least need to strive towards.
Benn Stancil: Yeah. And I think that’s the, the, like you don’t recognize doing it is in some ways, the sign of success, as you’re saying that we are enormously data-driven in the way that we decide where to go eat, like this is, we have fundamentally changed how we choose where we eat and it’s all, data-driven, it’s all built on like what our Yelp reviews. It’s like, how many reviews do they have We look at Google reviews. We like do all this assessment of like, well, how new is the restaurant And how do I say, like, there is a lot of analysis that goes into that process, but we don’t think of it as such because it is so embedded in just like, you know, search for a restaurant. And I do a bunch of analysis, kind of like in my head about thinking about it. If I buy a ticket on kayak, it’s kind of the same thing. It’s a very data driven process, but it’s not, it’s not presented to us as like a dashboard for buying a ticket. Here’s the interface, click it. We’re going to give you a bunch of data as meaningful as you make that decision. And so I think, I think like ML applications are basically the same word where I agree with you. Like if a company, if a company is marketing themselves as is ML, they’ve already sort of lost sight of, of the threat, because it’s just like, tell me what you do. That’s better. Like, there’s, there’s my favorite completely
Dave Mariani: Agree.
Benn Stancil: My favorite ML product is I think a website called like, it’s like remove BG. I think all it is you upload a picture and it removes the background of it. So like, say you have like, kind of like me on this day, it would just get like a, like an image without the background and it works great. And I’m sure it’s using some sort of InMail stuff behind it. I have no idea. I don’t care like that. That part’s irrelevant. And so I think that’s a pretty simple application, but like, those sorts of things are the places where ML will be mainstream because we won’t know what’s mainstream, not because this is, you know, some fancy NL product that does an amazing thing that we all like Netflix, but with ML who cares, I just want Netflix, but better.
Dave Mariani: I love your comment about, if we want, if you want to decide where we’re going to go eat, we don’t look at a dashboard and it really is. We sort of trained ourselves, right That, that, that, analytics means dashboards and queries and reports. And, really it’s like, if we really want to get to using analytics for people to make decisions, those analytics need to be embedded in applications that they use every day, and help them to make decisions again. they don’t even know they’re using an analytics or doing analysis. The fact is they’re just, they’re using an application to, to, to achieve an outcome. and I think that that’s, if we can get to that point where we can embed analytics the same way we can embed ML, I think the world’s a better place and there’ll be a lot fewer vendors out there. and more application vendors and less technology vendors who knows.
Benn Stancil: yeah, I think that’s part of this is just the natural process to get there. Like I think it’s, you know, we’ll get there eventually. we’ll go through these hype cycles before, but okay. But I think that’s with you where we met.
Dave Mariani: So, so, so Benn, we we’ve been chatting for a while. I always like to, to ask, my guests to predict the future. So, and, and you can, you can use any timeframe you like here, but, but what are some of the trends that you’re seeing that you, that you’re seeing that you think, could change the way we think about data and analytics in the, in the future
Benn Stancil: I mean, I think, I think that there will be, I think there’ll be a general movement and, and things like basically what’s happening in, in the public markets and sort of forcing some, some of the tamping down some of the potential frothiness, of the space, I think could be a catalyst for this to some extent. I think that there will be kind of a, a bit of a refocus on some of these things that we’re talking about, where this is the kind of natural cycle of how these like evolutions go, where there is a giant phase of technology for technology’s sake, figure a bunch of stuff out. Let’s kind of like explore the territory a little bit. and then that you figure out what is useful and what isn’t, and, and for the most part, like there is a significant calling of the things that you do, because you are saying, Hey, this thing found that be useful.
Benn Stancil: Let’s really emphasize that as things get and let’s get rid of it. and sort of every, I think like technological wave, except for crypto, which has not yet figured that out yet, but continues to charge ahead. there’s, there’s, I think like will be some of that that will happen. And so I think that’s like a good thing for the maturity of the space and what it is that we actually solve and things like that. it also sometimes makes it more interesting because I think people will be more focused on building things like removing the background images, as opposed to do all the mail that could in theory, remove the background. the second thing that is like, there are part of that I think is probably the stuff also becomes more mainstream and the scent, like the yellow bit, where the data becomes more injected in the various things that we do in ways that we don’t necessarily notice, in the way businesses get run, the sort of operation, you know, if I’m a support person, I suspect there will be more in the way that I do my job, even though it may not necessarily start be serviced to me.
Benn Stancil: And like I’m always checking them into dashboards. so I think there was some of that, and I though this is like a kind of more insider baseball type of deal. I think like kind of curious what happens with the space and like consolidation basically. Like there’s a lot of just explosion of things. I think we’ve had a lot of, people trying to figure out exactly where they fit. And I don’t think that can hold. I think we’ll end up with a handful of basically choices across these are the different ways that you can do these things. There’s not going to be that many. that’s not necessarily saying like the whole market comes collapsing down. It’s just like, there’s, we have created a lot of problems that I think are the consequence of just being so much sort of noise in the space.
Benn Stancil: that I think really there will be some, some like, alright, how do we, how do we make this sort of more usable and understandable and solve all those problems instead of how do we make it, the technology itself. Like I have this sort of notion for one of the blog posts about, about a modern data experience, where the idea is like, okay, we put it on a technology with a stack. How do you actually use it What’s the experience of using it Like, and I feel like that’s where it will be. We’ll be more focused over the next five years basically is all right. We need to, we need to figure out the experience of this thing, not the technology of it.
Dave Mariani: I love it. I love it. Well, yeah, you, you, you sort of alluded to this, wrote about this last week, but you know, what’s your, what’s your take on lake house versus data warehouse Does it matter Is it just semantics you know, it’s, we obviously have two, two big, competing vendors out there who are trying to encroach upon each other’s space. Do you think that, do you think it matters, Benn
Benn Stancil: I think it’s mostly semantics. I mean, I, I don’t, I think they are, they’re ultimately be the same thing, which is probably more of like the lake house, I guess to me, it is like, it’ll be more architecturally, a lake house more. Experientials a warehouse where technically what it is is it’s like a giant S3 bucket, I guess, of a bunch of stuff that you have different entry points into that you can query. And like one of those entry points will just look like a database. and I, there was an interesting post, was this, it might’ve been Natty. Somebody from DVT may have written something like this. This was a while back. That was about what it was about. Like whether or not you could have one warehouse to one database to rule them all they could do sort of productional stuff as well as analytical stuff.
Benn Stancil: And his point was kind of like, it depends on your definition of a database, where no, you can’t have a database that does everything. If everything is the same, if, if the database has to be the storage and the compute and kind of the whole thing tied together, but you probably can do it if like a database is a smaller piece of that is what was like a snowflake, a database, no snowflakes, probably just a product. Like it’s a product that has a database interface, but it’s not really a database. It’s just a product. That’s the word data. And, and so I think like to me, that’s the lake house warehouse thing is these are just products, they’re products that are going to look kind of like databases, but they’ll also have some other interfaces. The distinction between the two is isn’t really that meaningful. It’s, they’re all gonna kind of move in the same direction of like kind of this one database to rule them all where you can supposedly do a lot of different things. So
Dave Mariani: I like your, I like your concept there of entry points, because you think about S3 and S3, just, just, data storage by just raw object storage is really, it really has lots of different entry points, right You can, you can run queries against it with something like an Athena. you can use a product like snowflake, which is basically using as 300, the a hundred, the surfaces as a storage layer. you can run a spark on top of it and do your metal jobs. It really is the central storage that has many interfaces for doing different tasks. So I really liked that concept of if it’s transactional, if you need transactions and you’re going to have a transactional engine, if you need to have, an analytical, quick set of queries and you might have an analytical interface, but ultimately that data is still the same. That’s not how it is today. Right. We actually take that data and we physically load it in proprietary formats, but I’m hoping that we come to a day where we can, we can write the data once and read it many with, whatever the appropriate engine is for the task. so
Benn Stancil: Yeah, and that, that’s, that’s kinda my, in the most recent thing about Databricks that’s basically to me, what Databricks should be marketing is what is it It’s a giant bucket that’s kind of, you can put whatever you want in it. And then we have a bunch of engines that sit on top of it that you can treat it like a database. If you want, you can treat it like, you know, Python environment, if you want, you can monitor against it. Like it doesn’t matter. All it is is a bunch of interfaces into one giant sort of endless bucket that you could have the same. Like you’re sharing that data as the same in each. And regardless of the interface you interact with. And I think that’s pretty compelling. I don’t know if that’s a database or not. I don’t really care, but it’s probably going to use data.
Dave Mariani: There’s a database engine, but it’s not necessarily a database. I mean, I think that’s really where we have to sort of change our way of thinking. Right. As you think about the API or the entry point, as opposed to, the actual thing itself, cause we, we definitely conflate the storage of data with the processing of data. And, we’re obviously seeing the big ideas to separate the processing from the storage. And I think that’s probably the right successful model going forward, when it comes to architecture.
Benn Stancil: Yeah. That, that, that feels pretty clearly. Right. And I think that’s the first step of that is, you know, separate the processing from the storage, but make the processing all kinds of what it used to be. And now it’s just like, all right, let’s add new ways of processing it for new interfaces for that makes sense.
Benn Stancil: The lake house is warehouse. I don’t know who cares.
Dave Mariani: Yeah. I love it. I love it. Yeah. Shareholders. So this is like, this is awesome. any, any closing thoughts for the listeners, that, that you might have from the you’ve you’ve, you’ve, you’ve dropped endless amounts of wisdom on this audience and me so far, but, anything, anything they should keep in mind
Benn Stancil: nothing, nothing comes to mind go Hornets.
Benn Stancil: Yeah, we got one game left. We’ll see how it goes. but no, I mean, I, you know, I think it’s a lot of stuff obviously is changing. it feels like it is a time that we’re starting to feel a little bit of things to settle, but like that’s where we’re, we’re the end of the beginning. It feels like to me, where we had the giant explosion of stuff, and now that’s, it’s a long way from being settled, but like the questions that people seem to be asking are how do we settle things down, instead of how much further do we go and so I think that that to me is kind of the marker of, I don’t know where it’s gonna land, but, you know, we we’ve rarely apex and we’re all kind of falling back down now to something more coherent. we’ll see where we go.
Dave Mariani: It’s a sign of maturity, that’s for sure. So, that’s Benn, this has been an awesome conversation. Thank you so much for joining the podcast and, and to all the listeners out there. Thank you for joining and, stay data-driven.
