The Data Stack Show - 204: Will a Duck DB-Like Excel Emerge by 2075? And Is Data Every Company’s Most Valuable Asset? Featuring Benn Stancil of Mode

Episode Date: August 28, 2024

Highlights from this week’s conversation include:Benn's Background and Journey in Data (0:48)Reflection on Strategy and Vision (2:10)The Importance of Doing It Your Way (4:10)Early Experiences and Blogging (6:27)Self-Imposed Pressure in Startups (8:24)The Challenge of Decision-Making (12:11)Key Decisions in a Startup's Trajectory (15:48)Understanding Startup Anxiety (17:24)Importance of Focus in Data Startups (20:02)Product Market Fit Insights (24:38)Cultural Change and Product Fit (30:23)Evolution of Data Teams (31:57)The Role of High-Profile Data Successes (34:12)Challenges of Data in Smaller Businesses (36:16)Product Team Dynamics (38:18)The Future of Excel (41:11)Anti-Patterns in Data Usage (45:05)Imagining Excel's Replacement (47:05)Exploring New Data Solutions (49:24)The Role of LLMs in Data Analysis (51:29)Final Thoughts and Takeaways (53:10)The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Hi, I'm Eric Dotz. And I'm John Wessel. Welcome to the Data Stack Show. The Data Stack Show is a podcast where we talk about the technical, business, and human challenges involved in data work. Join our casual conversations with innovators and data professionals to learn about new data technologies and how data teams are run at top companies. Welcome back to the show, everyone. We are here with Ben Stancil. You probably all read his newsletter. I know we do, Ben. Welcome back to the show. It's

been years, I guess, at this point since you were first on. Yeah, awesome. Yeah, thanks for having me. I'm excited to be here. All right, so much to talk about, but just give us a quick background. Yeah, so I started my career at a startup called, well, I started my career in DC doing like policy research for a couple years, ended up at a startup in San Francisco, shortly after on a data team there. Shortly after that, I was acquired by Microsoft, worked at Microsoft briefly. Most of my career was then after that, me and a couple other folks started a analytics BI. We'll figure that out later, I guess, company called Mode, which sold away for folks and data teams to make dashboards and charts and do analysis and things

like that. Worked on that for about 10 years. About a year ago, we sold to ThoughtSpot, which is a larger startup BI tool. Was at ThoughtSpot for a little bit, and as of now, a few months ago, no longer with either of them. And so I am unencumbered, I suppose. Yes, unencumbered, unencumbered, yes. I swear to God. All right, so Ben, for the show, we started talking a little bit about Excel

and whether we think that will exist in 50 years. So that'll be a fun topic. You know, AI was part of the conversation there and just how we work now. But what are some topics that you want to dive into? Like I said, I am now unburdened by... Say whatever you want. Yeah.

To the credit of both the Mode and ThoughtSpot marketing teams, they were never emphasized a whole lot of editorial oversight. But yeah, I mean, all that stuff is, it's like a lot of stuff going on but yeah i mean all that stuff is it's like a lot of stuff going on interesting to kind of to to see where it's headed and get y'all's great well let's dig in ben so excited to chat with you and what an interesting time to you know recently you you sort of ended a decade-long journey building mode and sort of began a new journey with mode as part

of ThoughtSpot. You've written, you know, a lot about that. But I thought it'd be really interesting for me personally, but also our listeners just to reflect on that a little bit. And so I'm going to position you as an advisor to your younger self. So this is the completely unfair and unrealistic, like, you know, hindsight view. So, but also I have a couple of questions, but the first one is around sort of strategy and vision for a company, which over a 10 year period, I think is a really interesting thing to think about because that can change so drastically for a number of different reasons. And you've seen really a space sort of emerge and change drastically over that time period. The underlying technologies change drastically. But if you had

to go back to younger Ben as an advisor and say, okay, from a strategic perspective like here are the most important things to keep in mind what do you think you'd say and i want to add one quick thing to it i because this is very much related i'm really curious to on if there's a focus of like hey i wish i'd spent more time doing like x like i spent more time with this team or develop this team faster, whatever. I think that would be a really nice component here. Good one. So, okay.

So there's a handful of ways that answer that though. They all kind of have the same theme. Probably. There are a lot of things I would tell my past self and most of them are actually kind of framed around one rough idea, which is like, you can do whatever you want. I remember early on, I was talking to an engineer. So I'm not an engineer. I've learned how to do a little bit of stuff between now and 10 years ago. But I'm not an engineer.

And so I was talking to one of our engineers very early at Mode. And I remember him telling me, I was asking, can we do this? Or is this possible to make this thing? And he was, his answer was essentially, it's a computer. It can do anything. It just depends on how long, much time you want to spend in doing it. But like, yeah, we can make it do anything. And there's like an element of that in running a startup where you kind of feel like, okay, there are things you are supposed to do. There are rules. There are ways that this has to be done. If you're doing something, it seems like it's working, but you're not doing like the stuff that the textbooks say you're supposed to go do the stuff that the textbooks say. There's a bunch of stuff that you find yourself, especially when it's the

first time you're doing it, which it was the first time all of us who had started Node were doing it. You kind of run the playbook that is the thing that the internet tells you to do. And I think most of the things I would have done differently are like, tell myself you can do this the way that you want to do it. And so there's two examples of that. I think one is early on, and this actually lasted for a long time. I remember it feeling like weird that people joined to work on it. That I was not very old. I wasn't terribly experienced. And like, we had some really good engineers join early and some really good folks in the marketing side and design and all that, that I was kind of like, why are you here? Like, what are you doing? I could never quite shake the feeling that they

were like doing me a favor. So on one hand, I think that's not terrible. Like, okay, great. That means you're not going to be like an evil tyrant. I'm sure I was an evil tyrant in plenty of other ways. But like you, you try not to make them all mad. But on the other hand, you end up sort of realizing like you don't, if there are things that you want to do, you kind of be like, you have to have permission to do it. And I think one of the things that, that have done, me personally and as a company more generally, is like, hey, we want to do these certain things.

We believe this is the thing we want to build. We don't really care what other people say. This is the way we're going to do it. And so one example of that is, and to John, to your question, what's something I would have spent more time on? When we first started mode I had nothing to do like there were three of us the Derek who was our CEO was personable and could talk to investors and he was like a good face at the company so he went off and did those things Josh who was our CTO was chained to a desk building a product and I wasn't capable

of either this I was like not the person we would have put in front of investors I was certainly not capable of building product and as well as I kind of had nothing either of this. I was like not the person we want to put in front of investors. I was certainly not capable of building a product. And as well as I kind of had nothing to do, like my background was as a data person. And so I started writing a blog. It was not like the, it was essentially like analysis of public data that the very first blog post on Mode's website, which I think is still up, is this analysis of Miley Cyrus and the VMAs. Like, I don't know what it's about. And so I did that like partly because I was interested in it.

It was like kind of fun because it was like, all right, this is a good way to just sort of talk about data things. And it sort of worked as like content marketing because it wasn't like, here are five things

that you should do as a data team. Number five is buy mode. Right. It was more like 538 style stuff and this was actually pre-538 being what 538 was just like a nate silver blog on new york times at that point data-driven journalism and so like it had a bit of an like some people started reading it had a bit of an audience and eventually it became like okay now we have customers now we have other things to do and we looked at that as like, all right, well, this blog is not the right thing to do.

This clearly isn't, it's not in the playbook to write a weird blog. Like, therefore I should go off and do other stuff. And so I did. And I think that was a mistake. I think it was a mistake. It was like, we found something that sort of works. Just keep doing the thing that works,

even though it doesn't like feel like the right stuff. And you can find a hundred examples of that, I think, of places where you are reluctant to do things the way that you think you should or that it works or whatever, because there's kind of like a, that's not the right way to do it. And so I don't know, I think I would have basically wanted to tell myself like, hey, do this the way that you feel like it works and don't worry that much about like whether or not it fits the sort of playbook or not. One question on that. Well, I'm sure there, I will have more questions on that on this because it's so fascinating, but how much of that pressure to follow the playbook do you think is self-imposed? Cause I've felt that as well, right? You know, where you you i felt the same thing starting a company you know it's like there are these really successful people you know even especially people who have like serial

success who like write about these you know patterns that they've seen or that they've experienced or applied you know and so there is that pressure. But then there are also other, other pressures, right? Like, if you raise money, you know, I've had experiences before where it's like, okay, well, you know, these are successful investors, they're looking at these patterns and sort of this playbook. And so there can be pressure to like, do things a certain way, or, you know, sort of model things in a certain way. So in your experience, how much of that was sort of internal and self-imposed versus like sort of external? I think it's mostly self-imposed, but all of the, like you're on to me a gentle slope, like the external slope is gentle, but pointing to pointing you towards the playbook.

And so you do have to resist a little to your point about investors investors are going to mostly give you kind of the standard here's the way to do stuff kind of thing and yeah so it's not going to be if you try to do something totally crazy they might get on you and stuff but like they're not going to be that most of the time obviously there can be exceptions not going to be that like authoritative about like you have to do it this way yep honestly they become more of that later stage as like when they got yeah but early stages are typically not that hands-on anyway yeah but there is like a gentle pressure there i think to me the thing that like the way i would frame it is startups businesses in general are basically like this factory that produces money and fame and success and all that stuff and as a person who runs one of those businesses you're essentially

sitting there with like a thousand levers in front of you that are somehow connected to this factory that produces money right i love this mental image. For a startup, you have a thousand. I'm making minions. That's my mental image of like. And there's wires going from the levers back into the engine. And the wires are all in this giant knot. And some of them you pull it and it takes an enormous amount of effort to pull, but you can't see what happens after you pull it. You can't see if like, maybe there's electricity flowing through the wire

really slowly. Maybe the electricity turned into a fuse and it's about to blow it up. Like you have no idea what's actually happening. Like you basically pull up, sometimes something happens immediately, sometimes nothing ever happened. And to me, that's a really hard,

like that to me is why startups are hard. It's like, yes, there are a lot of work, but it's not, I want to be like, not to say that being an Olympian is not hard. I obviously am not an Olympian. I'm not in this very hard, but there is part of me that sees like to be an Olympian, you have to work incredibly hard, but there is a little bit of a path there. You have a trainer who tells you what to do. And obviously they may be wrong. You have to choose a trainer and stuff like that. But like, there is just, I have to put in the effort. I have to put in the hours. It's a tremendous amount of effort and hours, but I sort of know at the end of the day, if I get there,

if I don't, it's because either I didn't put in the time or I just like, I wasn't able to do it. I just don't run fast. For a startup, you don't know. You just don't know what to do. That most of your time is like, the business isn't going the way you want it to, or something's hard, something's working, something's not, whatever. And you're like, how do you fix it? And you're like, I have a thousand levers and I don't know what to do. Yeah. So this is, the context here is really funny, right? Because this is a data startup, right? So like a lot of the point of your tooling is to give somebody insight into their data so they know what to do, right it i mean but of all companies like you would think like oh they're a data company they've got a lot of smart people they've got

analysts they've got like way more resources than me but like it's still true of a data company like you know silicon valley you have pretty good access to talent like all that's true and it's still the same problem right yeah yeah and especially for an early company it's just like yeah like one of those levers is do we build giant feature x or giant feature y like i have i don't know i don't know how that's not how that's connected to the engine and when you're a big company like it's still hard and you don't know what to do but there's a lot of you've untangled a lot of the wires yeah you figured out a lot of those things yeah and for startup you have it and i think

back to the sort of original point of like just have some confidence in doing what you want to do. In a lot of ways, I think I'm a lousy founder. Like I'm just constitutionally a lousy founder because you can react to that confusing set of levers and engine by like carefully pulling stuff and being like, what happened? What happened? And the answer is like, nothing ever is going to happen. Like the thing is just going to do a bunch of random stuff. And you're going to do that. Right. I think the only way you really solve that is you're just like, you know what? I really believe this is the lever that it's going to be. I'm going to pull this lever. Like outcome be damned for two years.

I'm pulling these levers. I know that's what it's going to work. Sometimes you miss, but I think you have to have like that kind of delusional commitment almost. You know that what these levers and wires do and some people do and some people don't. We're a bunch of analysts. Like we, we pull the lever slowly and try to figure out what happened. It's such a complex relationship between did this work because I believed it would work so much.

And like, you know what I mean? Cause like, it's not like a linear thing where it's like, Oh, there's one right thing or three combinations of right things. Part of the impact of it, I think from a founder standpoint is like part of this working is that people can tell that you believe it's going to work. So your team's behind you and your customers are like, man, I believe it, you know? So it's not, it's not even like something completely determinable via analytics. Yeah. Yeah. And I think in a lot of these cases, like the decision doesn't, the lever you pull doesn't matter that much. So long as you're pulling it.

Right. Like you could build a lot of products that work. You could build a lot of things that, yeah, eventually some things may be terrible ideas, but like a lot of things, if you sit around and think about it for a week, idea one and idea two probably are both fine.

Like the thing that you need to do is just commit to one. And I think the challenge is if, if you are analytical, really, and you spend a lot of time like trying to split the tickets, and that's like that time. Whatever happens with like which lever you pull, the one where you're just constantly fiddling with everything, that one doesn't work.

Yeah, yeah. That reminds me, John. So John, actually, we had a chance to, we were part of a group actually at Ruddersack and we were doing this kickoff and the former CFO of Atlassian, Alex Estevez, who's a, you know, he's, you know, sort of famous for a number of reasons, like really well-known and super successful. And he came and gave a talk and we did a Q&A, which was awesome. And one of the things he said that you kind of know in your head, but really echoes what you were saying then, was he was like, okay, in this startup, you make thousands of decisions every year. And he said, but really, in terms of the trajectory of the company, there are probably like one or two key decisions that actually have a material impact on the trajectory of the company, there are probably like one or two key decisions that actually like have a material impact on the trajectory of the company, you know? That you may not know at the time. That you may not know at the time.

You could never have known. Yeah. Right. Right. Yeah. That is, yeah, that's really fascinating. Yeah. And I think that's basically like, it's tough because some things you, a little bit contradictory

to say this based on like, just pull a lever and commit to it. But also part of it's like, you just have to try. You have to just like, try to pull the thing and see what happens. And it doesn't mean just like willy nilly yank everything all the time, but you can't stare at the knot and figure out where things go. Yeah. Like you got to just start pulling stuff. Yeah.

Not all the time, but like you got to start pulling stuff. Yep. I think another thing, the Olympian analogy, I think is really helpful because the other thing is you understand the inputs, but you also understand a lot of the benchmarks, right? Like your coach has studied, like all the major athletes have studied, you know, all of the mechanics and the physics of like how to compete and like body composition and all of the mechanics and the physics of how to compete and body composition and all of that sort of stuff. But when you're trying to build a startup, a lot of times you're trying to build a novel approach to solving some problem. And so there are also very few benchmarks. You have this tangle of stuff, but then also you, even if something you're doing something right,

it's hard to know. It's hard to feel that in a visceral way because it's like, well, we're trying to do this differently than a lot of other people have done it before. And so like the physical sensation of a benchmark or like comparing yourself is almost non-existent as well, especially in the early stages stages i think i i think that's mostly true the one semi-caviot add to that is i do think that there's a lot of startups this is particularly when things are struggling a lot of startups that actually like you do have some feel like you have a part of it just like anxiety but you kind of can tell when something doesn't work like yeah you often will feel it before it shows up in anything else you'll kind of realize like how you talk to these customers

and nobody really seems that excited about it or like just pulling teeth to get this thing to work like i think there's a lot of a lot of like the job as a founder is this sort of delusion but but you also i think have to pay attention to like where are you in the dark moment like not dark in terms of bad but like you're lying in bed at night how does it actually feel and if it's like i feel something that's not good or i'd be like hey this is there's really something here i like feel a spark or something like there's usually something in that that i think is like it's not saying hey just like go with your gut on everything, but it's that your gut tells you something.

And if your gut's telling you something, I think there's a lot of startups that, that know it doesn't work for a long time before they like actually really make the drastic changes they try to do to like change something. There's a lot of, I kind of can tell this idea is not going to work, but I'm like fiddling with it for too long because i mean you get attached to it it's hard to change that but like most of the time when people shut something down or realize i have to pivot they knew a long time before that happened that like man yeah something wasn't right here yep man that's yes yeah that internal benchmark of like, yeah, that's, I think those are such wise words. Well, shifting gears just a little bit, you've obviously spent a lot of time

thinking about the data landscape generally, right? And so in terms of something working or sort of a data product working really well in the current landscape are there like a set of ingredients that you think map to the landscape that are sort of required in order for a data company to be successful in the current landscape like if you were starting a data company or whatever, like what's the inputs that make it good? Yeah. Or even like the inputs that make success like a viable possibility, right? Because there's so many things you can screw up or, you know, or things that don't work out. But like, are there any core ingredients that you think are just mandatory? I would say like, I mean, nothing's like really mandatory. I would say there's a handful of things that I would like feel better about.

One is related to this, like kind of have an idea, commit to it, know what you're doing, the way I'd frame that with like a data startup, it's more like know your boundaries, that the data landscape and like the tooling landscape is really overlapping. Like there's a ton of stuff that's all mashed up together where it's hard to know what bleeds into what. is really overlapping. Like there's a ton of stuff that's all mashed up together where it's hard to know what bleeds into what.

And so like BI and what mode the space mode operated was very much like this where we were a SQL IDE. There were notebooks attached. It was kind of data science. It was like mostly analytics. It was also kind of like

capital D data science of people building models that obviously bleeds into dashboarding and BI, which bleeds into things like alerting and then starts to touch on things like data discoverability and how do you like monitor things and there's you end up being able to basically be adjacent to everything right and so i think it's really hard to build something not knowing where like exactly you are saying this is the line that we will cross because your customers will pull you into all of those adjacencies. And so part of it is just like a clear sense of vision of where you what you want this thing to be and being disciplined enough to stick to it. Because, again, you're going to get pulled in those adjacencies.

There's always going to be like a big a big customer, if we only build this one thing, and I think that the real thing that happens there is you end up, especially early, you end up building a core set of features that everybody buys. And then one customer is like, we really want this one thing. And you're like, okay, we'll make that. And then other customers, we really want this other thing. And you kind of make that. But you will tell yourself you've built the same product for both of them. But in reality, you've built two different products where like they believe your roadmap is moving in their direction. And so like you've actually sold nine different roadmaps that way.

And so you have to have a lot of discipline of being like, no, look, that's we're not going to build that. Not because it's not adjacent to what we're doing, but because we know where it's going to go and it's going to be this huge surface area and stuff like that. And like there. And like, there's again, so much like space in the ecosystem of just like so much sort of ground to cover that I think you have to be like, we're going to be really good at this particular piece of it. That's what we're going to do.

And if we're either going to live or die on that piece being the thing that we think it can be. Is there a partnerships play there too? Where you're just like, where like, if you have a, like more specific boundaries, does that mean, does that be think it make easier

to have like partnerships? Like I'm thinking about like BI tools with a data warehouse or even inside a BI, like, well, we don't really do notebooks, but like these guys do notebooks well. Like is that, because, you know,

a large company is like a lot of them have a, you know, a bunch of different BI tools. It's not like they just have one. Is that a part of it or is it more just like about the focus and like that doesn't really matter? Like informally, sure. I think that's fine. I don't know that. I think partnerships are often a distraction for folks.

So there's a couple of reasons for that. One is that a lot of times people will be like, look, we want to be this thing that works really well with the ecosystem. The ecosystem doesn't want to work with you. Like, they're the incentive for them to actually work with you. If you get big enough or popular enough, that changes. If you're Snowflake, if you're DBT, if you're Fivetran, if you're Databricks, a handful of others, like, yeah, you can kind of, you will have the gravity to have people come work with you, but you're not going to get there from day one. There's no like, oh, we want to have partnerships with everybody from day one. Sure.

Plus you can't build good experiences with those things. Like, I don't think you can say like, hey, we're like, you can't be like, we're going to be great with all these database partners. You're not going to make a good product there either. I think the only way that partnerships really work is you choose like one person that you're saying, we are absolutely going to ride these people with coattails. Yeah. That we are going to be basically an extension of snowflake or data bricks or

some amazon thing or gcp or dbt or whatever yeah where you want to be so good with that thing that their sales team sells you yeah but the idea of like, we'll be on some partnership websites, that's a waste of time. Sure. Yeah. Yeah. Yeah. It makes total sense. So kind of related back to the conversation or back to the point you made about knowing, you know, if something's working or something's not going to work and sort of feeling that deep in your bones, speak a little bit to product market fit. Like how would you, even thinking back on your mode experience or just your current thoughts on that,

there's so many thoughts on product market fit, but just interested, again, thinking about like building a company, you mentioned focus, like how do you think about product market fit? It's a feel thing. I think it's a lot of, it's just like how hard is it to sell

and who can you sell it to? That there's not, you know, there's not like a, you don't have it. There's a lot of people who've said this before is it to sell and who can you sell it to that there's not you know there's not like a you don't have it there's a lot of people have said this before it's not like you don't have product market fit you do you sort of it's like you have a product market it is what it says you have a particular product that fits a particular market and basically i would like define do you have a product that fits a market is can you sell it pretty easily do people come to you to buy it like the feel part of that i think is actually like if you're on sales calls it's kind of like how excited are you to join the sales call you know you're about to give a demo to

somebody are you like dreading it because you know you're about to just get hammered by a bunch of questions that you can't answer and they're like you're gonna have to disappoint this person over and over again because you know you can't solve this problem you're like hell yeah i'm fired up for the sales call because i know i have a thing that they're going to be excited about i'm excited to see the look on their face when they see this thing i think that's almost a better like yeah okay fine look at retention look at grow whatever but i think like that feeling is almost as good of an anything as measure of anything because you know how you feel going into these calls you know how you feel when you're like getting on a call with a customer. If you're like,

God, like every time these things happen, they're upset about something. Or if like you get on and they're like, I know they're going to love it. I know they're going to talk about how great it is.

Yeah, they're going to have some things they want differently or whatever. But like they're excited customers and they're excited to be on this call because they want to make this thing better together

versus like they want to yell at me about why they hate it. And so I think like product marketing because in some degree just that feeling it's but i think the thing that you you have to be careful about is like that doesn't just mean like hey we have product market fit or don't it's like oh we actually work really well with this type of person but not this type of person or whatever and so it's can you is that type of person big enough i think one of the things for instance like mode the place where mo like one of the reasons we actually ended up doing the acquisition on a much longer time horizon basically was the market for a data tool, like an analytics focused data tool for technical folks isn't that big. It's just not that big of a market to sell a productivity tool for analysts it's why all the tools that do

that end up kind of becoming bi is because like the only way you can either get there is either sell it for something that's really expensive which people now are kind of allergic to like nobody wants to pay five hundred dollars a month for software like that just doesn't feel right anymore they want to pay 25 and say like oh we're going to sell 25 seats well i guess we got to sell to everybody and so you can have a product market fit with like the analysts where you sit on calls with them and you're like, it's great. But then you get on a call with the head of marketing and they're like, how do I do X, Y, Z that I'm used to on my BI tool?

You can't do it. So that's where like, I think you have to be honest. We're like, yes, we have product market fit. Doesn't mean we're set. It means like we can sell to this particular group. And how easy is it to find them how many of them are there that kind of stuff like where actually is the draw yep so i think i have an interesting perspective on this because my um previous company we bought we did a whole like

selection process for bi tools and went with mode around 2018 i think and i and there are some things that like since then i've realized were not typical of the company and one of them was the analysts were very much involved in the process which is ironically I do not think that is often true in the selection process and we compared a lot of tools and absolutely like the analyst like at the time that was involved in the selection was like, this is great. These people get me. This is how I want to work.

That was a big part of the decision. And then we had a lot of, what do we call them, civilian analysts. A lot of dual role type people that picked up a little bit of SQL. So it just worked really well in that environment. And since then, I've learned that most companies, this is not you don't like we had a sales manager that was pretty good at sequel and would like use mode like that's not super normal but on the product market fit side i'm curious you've got two components so the product part and the market fit part which do you think there's over indexing on one of those where like people are typically they keep tweaking the mark

the product part and like we want to cram into this market and it doesn't make sense where the other side we're like we're in search of a market for a product that maybe doesn't exist. What angle do you think do people over-index on? Real quick on the selling to y'all. You're right. That's not that common. It's not that uncommon.

But this is one of the reasons why selling a product like Mode is to get it really big is hard. It's very hard to identify it's really hard to figure out which companies are like that and and so you end up having to cast away like you can't have a very targeted net because you're basically talk to everybody and then figure out based on conversations like how much say the data team has um there's not a good signal. It was not like a Facebook ad target. Right, you can't get an audience. Yeah, crap.

Data team influence, whatever. But that was very much like the thing for us was like, if the data team had a big voice in the buying decision, we were great. If the data team did not, then it's a struggle. So you're like, do you over-index on the product or the market? I think that most people mess up. Like they either don't interpret the market

as, like, they basically over-index on their own experiences. I think it's fine, like, that's all great, but they over-index

on, like, how common their experience is or, like, they really think this thing is great, the market will love it and there's just,

like, not that many buyers. Right. The other thing I think that a lot of companies do is they're selling cultural change,

too. Like, the product isn't just a product. It's a new way of doing things. And like, you can pull that off, but you basically pull that off by building something that just has such

good product market fit that like people adapt around your thing. Like DBT was, DBT pulled off cultural change. And to some extent that was the intent from the beginning but it wasn't successful because it was intent from the beginning it was successful because people just like they built a product that people adopted like crazy and so a bunch of people changed it to sort of fit their way of doing stuff around it but there's a lot of companies that are like we're going to come make you data driven or whatever we're going to change

your communication into some now you're going to be a transparent company you're going to come make you data driven or whatever. We're going to change your communication into some. Now you're going to be a transparent company. You're going to be a company that runs in new ways. It's like, yeah, you're going to make it those things. If you make the product just so appealing that people start using it and your product encourages that behavior, you're not going to make it that way because people are like, I'm adopting your product for a new cultural thing. Like there may be some buyers who say that, but they're not. That's not going to be, you know, like, like Zoom can't go to people and be like, we're going to make you a first company,

like adopt a remote first culture with Zoom. And they're like, yeah, you're right. This is a good tool for making us remote first. It's like, no, you're gonna make a good product. And then people will be like, actually, we can kind of be remote because this product is so good. The, you can encourage it, but like, I think there's a lot of data companies that, some of your points about like,

what are data teams doing, stuff like that. Like, look at it, like, we need to also encourage cultural change and it's just like, you're selling then two really hard things

and not just one. Yeah, sure. Yeah. Yeah. Speaking of data teams, on the, like,

data function side, I'm curious, like, how you've maybe seen over the last, like, 10 or so years, that evolve. Because we talked before the show, data teams haven't always existed, right?

And then in the last 10 years, I think there's been some evolution. But curious how you have seen that evolve, I guess, basically with Mode customers or just in general in the ecosystem. It's a little weird. So when we first started selling Mode, the first pitches to it to VCs, we were like, look, we're building this thing for data folks. They write SQL. This is what they're going to do. And we got a ton of responses that were just like, what are you talking about? Who are these people? These aren't people. That is like, we have to prove the market for SQL. We have to like explain what these analysts are. Like, are they data scientists? Like we know data scientists are a sexy job. Are you building for

that? And we're like, no, not really. And they're like, well, why aren't you building for that? I, and so like that has completely gone away. The idea of semi-technical analysts who write SQL and are trying to answer questions, and it's not just building dashboards, but it's not proper data science is a pretty standard thing now in a way that very much was not when we were first getting started. Obviously, a lot of stuff with transition to cloud and stuff like that was also early back when we started, where a lot of people were like, I don't know about cloud data tools. And now that's a snowflake hacks notwithstanding, something that most people are pretty comfortable with. The thing that I think is like, I can't decide actually if that's a good or bad thing that that's happened.

Like in some ways it's like, great, this is amazing. We have all these people who are going to think about these problems and be analytical and business are gonna be way better but like we haven't done that really like are we off making better decisions not really like we still are in the same still talk about things in the same way of like how do we get out of making dashboards and get on to the more impactful work like this has been the line forever and i don't know like i i struggle sometimes with like is there actually more impactful work there like we've had a lot of opportunity to do it we booked a lot of tools to try to make it more possible and yet we're not that good at it and so i don't know i think it's like there's been a lot of evolution around the belief that you give smart people big data sets that they will go and find smart decisions.

That has become like a faith that a lot of people have adopted. But I don't know if it's the thing that like the faith has delivered the goods yet. Why do you think people believe that? Like, where does that core faith come from? Because it is a non-trivial like it is a very widespread belief i mean literally entire companies like that make a lot of money are built on that i think i have two answers i think one is nate silver um like basically that there's a handful of very popular things that have happened where people are like look the data solved this problem

be it nate silver be it money ball be it sure a few things that are like high profile you know the warriors started or steph curry started to allow threes people were like the warriors really analytical all this stuff like you can point to a handful of examples where yeah analysis in some form or another seem to be very successful. Investing, renaissance, and the whole, you know, computers doing this and all that. Yeah. The renaissance is like, I mean, that's a decent example in some way. So the reason I think it's a decent example is like, the second thing I think that happened was it was tech companies.

It was a lot of like, it was Facebook. It was Google. It was Netflix. It was whatever and i think the thing with that which is also true for something like renaissance is they're solving problems that are like just fundamentally well suited for that yeah yeah this enormous optimization around like yeah google can make enormous amounts of money by optimizing ads on search results because there's a trillion of them a day or how i you know i don't know yes there's leverage there renaissance can make a ton of money because there is leverage in making a billion trades and financial markets that are automated based on like small little market signals yep

basketball and and money ball and like predicting an election it's like it was yard signs and now it's like let's be a little bit more scientific and there's something very useful. But that doesn't mean that like same method works for sure. We're a hundred person business trying to figure out who we sell. Like, it's just the data is fundamentally not that valuable. And so I think a lot of people saw the early successes in data where people who had data that was high leverage, like kind of obviously like that's why they were successful they had high leverage stuff and so then there was somewhat like of a kind of cargo culting around ah the data must have been the thing that like the analysis must have been the thing to have data it's our most valuable asset too and it's like it's probably not yeah yeah that

is super interesting yeah i think another interesting trend here that I've seen is you had like IT teams, for example, right? That had developers, say you had a product or say you didn't even have a product. You're a company and you have to integrate with other companies. You don't even have like a core like SaaS product. You're just doing integration. So you have a small IT team. And then from there, like I think before data was a separate team team it would be a business analyst or business user that goes to the it team like hey i need some data and like what do you need and then like you

know they pull some data down and give it to you in a spreadsheet and like okay so and and at least companies i've worked for like the data team started out of like it was done messing with it like they're like this is a waste of our time like let's stand up let's like hire a person or like stand up this data team so we don't have to mess with like, this is a waste of our time. Like let's stand up, let's like hire a person or like stand up this data team so we don't have to mess with this because it's a waste of our time. We're focused on our critical client integrations or our, you know, our SaaS product or whatever.

So I don't know that's true everywhere, but I think for a lot of like industry business, like in my like supply chain, like history, that was true. And I mean, that's not a, that's not a it's not a great start to a team right like as far as like yeah like i don't know like we don't really want to do that like you guys do that and then you can end up with some awkward like they don't have quite enough business context to be super useful they're not as technical as like this like formal it team and you can end up with some data teams in a really kind of weird spot.

Yeah. So I don't know if that's an industry trend or just kind of my experience, but... I mean, so there's a... The guy who used to run product at Mode is a guy named Nan. He now runs product at Linear, the, like, Asana competitor. Yeah. And he gave a talk at the, like, Figma had a big conference

I don't know how many months ago. He gave a talk at it. And the kind of core point of the talk was one of the ways that product teams really mess up is they have this kind of notion, and it're going to assign it a PM, two designers and six engineers. And this like squad of people is going to go off and build the admin functionality. And this other team is going to go off and build our mobile app. And this other team is going to go off and build whatever other thing. And his point is like, you end up creating a lot of symmetry because it looks nice. And because it sort of feels like organizationally tidy. But in reality, like that's really bad because probably the admin page is not as important as a lot of those other things but once you have a team that's assigned to it that's their job that's the thing they're excited

about they're going to figure out ways to like push it as far as they can because what else would they do yeah i think that you could kind of apply the same thing for some data teams where people like to your point about the it thing i I was like, this is a pain. It's hard data team to go get rid of and make it not a pain for us. And now you have these people that are like, we're going to push it as far as we can. Yeah, we want it to be great. It's how they interpret their own success, right? Like it's the lens through which they can actually, it's their barometer.

If you were on that team, that is the sensible thing to do. Like push it as far as you can. But it doesn't mean that like you're pushing something terribly important and so that doesn't mean like data stuff is important but i think it it it's very important for some like what's the business how is it performing kind of operating as like some degree of senses but i think there's a notion of like yes there's hidden gold everywhere. There are needles in these haystacks and it's like... Right. Yeah, it is interesting. I mean, I kind of think about data tooling in general,

and we have to get to the Excel question, but I work on cars sometimes and there's this sort of if you have the right tool, it can make certain things so much easier, you know? And so, but then you also see this huge over-indexing on like the tools, right? And it's like, well, actually like if you're good at it, you can just use a tool that's pretty good. And like, you just do the work and it's fine. And so to some extent, like data tooling in the data industry, like sometimes it seems like it can over index on the tool. And it's like, this is actually just a means to an end, right? It needs to be good enough to like, to your point, Ben, like help the business, you know, achieve like X, Y, or Z. Yeah.

Yeah. And the tooling stuff I get, it's having built one, you know, and there's like people who live in them. You want to make them good. You want to make them good. You want to make them. Sure. Right. And I, yeah, there's, I think it's fine, but it doesn't mean that it's producing something terribly useful.

Right. Yeah. Yeah. Yeah. Again, maybe that's okay. I having a comfortable chair at the office does not like necessarily make me way more productive, but it is nice. Yeah.

So yeah. Yeah. Yeah. Yeah. All right. All right. The Excel question, John. Yeah. does not like necessarily make me way more productive, but it is nice. Yeah, yeah, yeah. Yeah, yeah. All right. The Excel question, John. Yeah. I've been waiting this whole time. You've been waiting this whole time. Yeah.

So we're talking before the show about a little bit about Excel. It's been around. I don't do, what is the Excel launch date? Like roughly, like how long has it been around? Ooh, that's a good question. Get somebody to look that up. Anyways, so it's been around a long time. We're talking about, has it been around that's a good question get somebody look that up anyways so it's been around a long time we're talking about will it be around in 50

years and we also talked about as far as data teams is like there's at some point where almost no one or no one had a data team right and then at some point that happened and then we were talking about with ai like does that impact data teams going forward do they look different do they you know are they more decentralized kind of embedded into the business are they more centralized yeah so that's kind of a starting place but yeah ben just curious your high level of thoughts on excel first and then we can dive into some of the other stuff yeah i mean i think the excel stuff looks like it was released in 1985 okay wow i my my like question about the excel thing is yeah it's been 1985. Okay. Wow. My question about the Excel thing is, yeah, it's been around, I guess, for 40 years.

What, if we imagine a world in 50 years where it's not the thing that we all use, what does that mean? What does that look like? How do we actually get away from it? I think that the part of reason I ask is like, there's a lot of companies that attempt to kind of replace it. Like there's a lot of companies that they're like, their is excel users like look excel people they're doing all this stuff wrong don't want to do this and i think it kind of misses like there's a lot of anti-patterns in excel that are actually the reasons that people use it and i think there's a lot of effort in like sales companies to like solve the things that are wrong with excel that sort of miss that that's the reason Excel is good.

What's one example of that? Like one anti-pattern? Versioning. That there are tools that are like, no, we want it to be connected to live data. You don't want to have to send around all these versions of things. And it's like, you ever sat in like a sales forecasting call? You ever talked to like, you want like the person who's sitting there and they're like looking at their spreadsheet of all the current deals in the pipeline

for that thing that you like randomly update all the time? You have any idea how upset they would get? Yeah. Like the thing that I think is useful about Excel

is one of the things anyway, is when I send you an Excel file, I have sent you everything. I have sent you a standalone product. It is the data and the view of the data on top. It isn't just a prism that is looking at something that could change. It is like the database plus the BI tool all in one. And the logic is completely transparent.

There's something really that is, it's not going to change. I can fuss with it and I'm not going to break anything that, that it's like this nice little package that all exists in one, in one thing. And again, if I have, if I'm a salesperson and I'm doing a pipeline thing, I can be like, yeah, I want this thing. So I have a file that I know is not going to change. And like, you could say, okay, well, what about Google sheet? That's I think half true because like people don't really use the Google, like the Google Sheets connected to a database is not the thing that people want to use it for.

For sure. They use it as like, oh, okay, it's nice that if we're collaborating on this thing together, yeah, I don't have to worry about like merging updates. There's not like a track changes merge that has to happen, but it's not so much like, oh, the input data is just randomly changing edges that you have to like very manually say okay i'm going to put something new into this thing and so it's just like a it's a really nice package it's a nice thing to have that the versioning is kind of the point and so i think a lot of tools are like no we got to get rid of that

we got to be a sass product we got to be all this stuff that's more modern. It's like, I don't know. A desktop app with files is sometimes pretty good. I think another anti-pattern that's even more fundamental is that Excel will primarily use to work with data sets. Each cell is addressable, right? I can freeform type into each individual cell and I see a ton of Excel documents where it's like, it's not a data set excel documents where it's like it's not a data

set like it's like i type this thing here i type this thing here i wrote a note under that i like it's it is a like no sequel you know it's unstructured sometimes structured as well but it has it's addressable down to the cell level it's not just a data set yeah there's a lot of business users that use excel in a non-dataset way. Whether it's just maybe there's one tab where they've kind of like typed in some things and then the rest of it's datasets. But I think that's a really common pattern that you don't really see in like BI tools. Yeah. And you can like touch it.

It's a thing where it's like you can get your hands around. You can understand what you're doing. There is a, I think, like, this is one of the reasons I think DBT is pretty good, is it's data pipelines that is, like, table by table. And for, like, okay, that has problems and it can create a lot of mess and be expensive or whatever if you do it badly. But there is a nice thing where you're basically, like, able to see each step of the pipeline and just, like, I can run it. I can see it. Did it work? Okay, I can go to the next one. I can see it.

And Excel is a version of that where, like, I can run it. I can see it. Did it work? Okay. I can go to the next one. I can see it. And Excel is a version of that where like I can go in and I can flex with things and I can play with it and I can understand it and like BI tools and anything that just sort of sits on top of database puts this kind of unknown abstraction on it. Sure. Yeah. Like, I don't really know what I'm manipulating. They're like, if you work with an OLAP cube, for instance, like an OLAP cube, it's just, it's hard to get your head around what you're doing. And like Excel sort of solves that.

Like, yeah, you can create it and you can create a pivot table and you like, you can sort of handwrite it the way that it did a thing. Whereas if you're looking at something, you're like, I don't really know. Am I looking at this right? I don't really know. Like there's just this, this like lack of transparency around it that I think Excel makes really nice. And to your point, like it seems broken because it's not structured.

It's not governed in the way that things should be governed. It's not version control. Yeah, that's what makes it nice. If Excel was sort of dethroned, how would you even think about what would replace it in 50 years? Or what would have happened, right? Yeah. Or like this cosmic data event happened. I haven't ever quite put this idea together, but my version of it basically gets replaced with a different version of Excel that uses the internet more smartly. Again, a lot of Excel that like uses the internet more smartly that it's again,

a lot of tools that are spreadsheet based BI tools that sort of claim to be the next Excel. Don't let you, don't let these sort of do this easy versioning stuff. They don't let you like share files around really easily. They don't let me like double click on a CSV and open it that there's this like, oh, it's this heavy thing in some way. I think like a tool that would sort of be interesting to me

is one that basically like has this kind of hybrid model where it actually, you just download a thing on your computer. It's a piece of software. You double click on a CSV, you open it up. But so the way that MotherDuck works,

which is like DuckDB, they basically building like a hosted sort of version of duck db yeah they do this kind of hybrid thing where you like you can run it locally because i've talked to me where you can run it locally but you know those are just like push things up to either resources on the cloud or there's a way to sort of sync so i have my version of mother duck database basically you have your version we can kind of say like actually let's read off of this centralized one but we can also pull down our local versions and futz with things there and stuff like that it seems like there's a version of that for excel that basically the aim is like make it

really fast make it so that solves the scale problem make it so that you can manipulate across like really big things um without giving up the like close, close to the metal, I can play with this thing and just open files on my desktop. Yeah. Yeah. I thought, yeah, when you were describing that, when you were describing that experience, I actually thought of DuckDB

because it has a lot of the same characteristics. One last question, because we're at the buzzer and Brooks is telling us in the Slack channel that it's time to land the plane. So, you know, you're currently unencumbered, but if you were going to go solve, you know, another problem in the data space, what or even just like build another tool, what would you like? What problem would you try to solve? I don't actually know what sort of a tool like this, the tool around this would look like. My general belief right now about AI, I hate this as an AI answer, but whatever. It wouldn't be a data set.

It wouldn't be 2024 if it wasn't. My general belief around what AI does is there's a lot of people who are basically like, okay, let's take AI and like let's make a better BI tool with it where you ask it questions and it automatically answers whatever.

Yeah, yeah. I think like the more interesting thing is like if we think about what like LLM specifically are good at, they can read really fast and they can tell you

what they read and they're good at like summarizing that. And there is a lot of like interesting information and signal in unstructured stuff that there's a lot of information and customer interviews or just like videos of the world and things like that. And that information is probably much richer and more useful, honestly, than like the data that we collect. That if you want to say like,

how do we make an intersection have better traffic patterns? I don't know anything about traffic. There's a lot of probably urban planning people that will disagree with all of this, but whatever. In my head, it makes sense. If you want to make an intersection with like better traffic patterns, one way you can look at it is like, you can look at the data. You can like instrument the intersection and measure where

cars are you have like some structured giant file of things and you do a bunch of analysis on it okay it's kind of hard or in theory you could have someone sit there and watch the intersection for a year and just watch everything and if they remembered everything really well they'd probably be like you know what i noticed some of these things maybe we should change that and like actually that second one is probably more useful in terms of figuring out something that might work than handing this data to somebody and saying like go spelunk through a giant data set and try to find the insight like right if you just watch stuff if your job is to improve an intersection i bet if you just watch it for 24 hours and you had this amazing memory you'd probably come up with a lot of ideas about hey maybe we should try this i noticed this thing that always happens i noticed whatever that i think is

and like that is what an llm does an llm reads stuff it remembers it like is able to have all of it in memory and then sort of find these little patterns and i think it's actually like enabling that isn't taking the video and structuring it into a data set it's just like bypass that entirely it's doing what yeah a user researcher with perfect recall would do which is they're not like trying to map everything to numbers they're just saying like look i watched i sat and watched people use this product yep for a thousand hours. And like, I saw some crazy things. You should know about those crazy things.

That to me is actually where it's what data is trying to get at, but in a very indirect and hard way. And it's like, the reason we don't do the other one is because it's too hard to do anything with. It's too hard to take a thousand hours of video of customer interviews and like actually make any sense of it because you have to watch it and you have to have one person think about it and aggregate it and they can't do it but if you didn't have that problem give me the interviews and and so i think like there's something in that to me of like what do you do if you're trying to say actually we think most of the

valuable information is in this unstructured stuff. And the tooling we should have is like, make it so that, yeah, I can just watch things or interview people or look at texts and find what's interesting. Do the analysis on the unstructured stuff and not by structuring it, but by like the way someone, they just observe. Yep. Fascinating. Ben, this has been so fun. The time time flew by but thanks for giving us some of your time thanks for letting us encumber you momentarily for sure thanks for having me on yeah this was fun

the data stack show is brought to you by Rudderstack the warehouse native customer data platform Rudderstack is purpose built to help data teams turn customer data into competitive advantage learn more at Rud ruddersack.com.