This week, investor Aseem Datar is possibly making app dev and product development history with his guests. We’ve got former Tableau CEO Mark Nelson and James Phillips, the former president of Microsoft who drove the creation of many of the company’s most successful products, including of course, Tableau’s number one competitor — Power BI. James was also just named the VP of Google Cloud, post recording this podcast! Mark and James are both part of the Madrona family, as venture partner and strategic director respectively. And today, we have the rare treat of them together sharing their learnings and experiences about modern app development, data and analytics, product-led growth, and scaling in the face of stiff competition. You won’t want to miss this.
This transcript was automatically generated and edited for clarity.
Aseem: Hey, everybody, I’m excited to be here. Actually, I’m super excited today. I have the unique pleasure of welcoming James Phillips and Mark Nelson as our esteemed guests. And you both led and built products that are truly unique and market-defining, so this is going to be an exciting conversation. And before we dive in maybe let me just hand over to each of you for a quick background and intro, and then we can go from there. James, you want to go first?
James: Sure. Most recently, I spent 10 years at Microsoft and developed something we’ll talk about today, a product called Power BI, along with a number of other things. It’s super cool to be on this with Mark, who was running my primary competitor at the time. Prior to Microsoft, I founded a number of companies, Couchbase, Akimbi, a company called Fifth Generation Systems. Spent some time at large organizations as well, Intel, Synopsys, and did a little bit of a detour, spent two years as an investment banker helping technology companies go public and execute mergers and acquisitions.
Aseem: That’s awesome. I’m glad you hit the compete question head on. Over to you, Mark.
Mark: So I’m Mark Nelson. I’m currently spending my days as a venture partner at Madrona. Prior to that, as you mentioned, I was CEO at Tableau, ran product and engineering before that. And I’m coming off of a long career, which I think just means that I’m old, with a lot of experience in the data space. Started off working for a database company called Informix, which is now a tiny little division of IBM, and then spent 17 years at Oracle, which is where I really kind of grew up, before moving over to be CTO at Concur prior to the acquisition, and then hung around for a while afterward. And then went to Tableau, again, to run product and engineering, and then became CEO sometime after the acquisition.
Also lived through the two biggest software acquisitions in the history of Seattle and so that’s a small claim to fame.
Aseem: So awesome. I’m just amazed at the fact that there’s so much wealth and so much knowledge in this virtual room today, and can’t wait to dive in. So let’s just jump into it.
One of the questions that founders often wonder about is, how should you think of building products from the ground up? What are key items that they should keep in mind, especially when they’re thinking about a new category or value creation in something that’s never been solved before? James, do you have a perspective on that?
James: I do. I think probably the most important thing that one can do is to recognize that putting a product out in the marketplace is your first, I think, real opportunity to get real actual feedback. It’s sort of the ante for learning. And so the principles behind that for me are, get in market as fast as you possibly can, because once you do that, you’ve got the lines of communication open, you can begin learning, you can begin incrementally improving. And I think that pairing that with a system that allows you to incrementally improve and to ship very frequently while listening is the key to taking that feedback that you just effectively paid for, by building a product and learning and turning and burning, if you will, and improving the product as you go.
So fast-to-market, and perhaps as important, fast iterative cycles to continually improve the product because you’re learning once you’re out there.
Aseem: Couldn’t agree more in terms of the faster iteration. Mark, is that something that you guys also had as front and center as you thought about category creation or building products? Or do you have a different perspective on that?
Mark: No, I don’t have a different perspective. I’ll add onto it, and it’s this fine balance of being convicted about the problem that you’re going to solve and the who you’re solving it for, and then the ability to listen as you find out where you were right and where you were wrong. And it’s this fine balance between having this super passionate conviction that you know what you want to build while listening and getting that real feedback. You don’t want to overbalance on either one of those, right? As soon as you stop listening, you’re doomed, and yet, especially for category creation, no one’s done this before, so you have to be convicted that this is a problem that the world wants solved and that you can solve it, just to get across that initial barrier.
And so having that delicate balance and delicate judgment, and then 100% on getting out there fast, iterating, and learning, there’s nothing like having it in the hands of customers to get the good news, bad news, and feedback that you need.
Aseem: James, I know you have a perspective on the five minutes principle. I’d love for the world to hear it. I’ve heard it so many times, but nothing like coming from you. So can you share a little bit of that in terms of how you think about customers and how do you think about value creation for them?
James: Yeah, we had this saying, “Five seconds to sign up and five minutes to wow,” which was shortened to five by five, and it was really sort of a guiding principle for how we built Power BI if we want to talk specifically about that for a moment. The goal was that any user should be able to show up, sign up to use the product within five seconds, and at the end of five minutes if they were asked, “What do you think?” We wanted to hear them say, “Wow, this is amazing.”
And for the first probably year and a half of Power BI, we literally, every single week, ran a 5-by-5 user study where we would invite people in, we would sit them in front of the computer, we would ask them to sign up, and we would ask them to use the product. We started a stopwatch. We stopped it at five minutes, and we asked for their feedback after that period of time, in order to ensure that the ability to discover, ingress, and begin using the product, or at least have an emotional experience with it, was positive. And if we got that right, then we could start working our way down the funnel and driving usage, et cetera. But it really started with that, “Let’s go find users, let’s get them excited, and let’s hook them.”
Aseem: It’s so funny, it reminds me of the early days of Visual Studio and DevDiv. One of the things that’s similar in this vein was the number of clicks it takes for people to sign up back in the day, and if you’re an engineering manager, if you have to pay in terms of headcount, you don’t get a certain amount of headcount in your team if the clicks are more than 10. And so such a powerful principle to abide by, especially as you’re building early on.
Mark, did you have anything in terms of dos and don’ts or principles in the days of building Tableau? And what were some of the goals you set for the team?
Mark: Yeah, for sure. This predates my time at Tableau, so I’ll talk about it, but I’m very familiar with how this story went, and I’d like to think that a lot of what James instituted was a reaction to what Tableau had done in the market, right?
Aseem: Well played, well played.
Mark: Because it was Tableau’s mantra and go-to-market motion, from the beginning, was… Again, we predated what is now called product-led growth, but it was product-led growth. There was no sales team. It was amaze the analyst, right? That as soon as they get it in their hands, they realize this is something they can’t live without. You had to be able to download it, use it immediately and immediately, again to the five seconds to wow, you had to immediately get value from that. You had to immediately see this as something that was beautiful, and you could not live without.
And so there was a maniacal focus. Again, it wasn’t sign up because we started off as a desktop tool. So it was a maniacal focus on download time, a maniacal focus on installation experience, a maniacal focus on what was that first thing that got you going and how beautiful it was and how beautiful that experience was. Because that was the lifeblood of Tableau early on, it was the first X number of years was inside sales. It was all led by the passion that individual analysts had about the product.
And then the other part of the equation was the community that then grew up around that product. That was the go-to-market motion for Tableau for a long time until it grew up.
Aseem: So true that it’s worthy to highlight that both of you built these massive businesses when PLG was not even a thing, or I don’t even think the term was around back then. And so the question that I think most founders have is, what makes PLG the right strategy for founders that are early? I mean, a lot of it is, as a first-time founder or an early founder, you are selling it. You are doing this sale, whether it’s to the enterprise or the SMB you are sitting in on, it’s high touch to a certain extent. At what point does that PLG equation, in today’s terms, come into being? Is it business-specific? Is it industry-specific? How should one think of PLG, and what makes it the right strategy?
Mark: I’m happy to dive in here because Tableau was an early poster child, whatever you want to call it, and then I had the fortune, being part of the Salesforce universe, of seeing Slack come in. And I think the modern definition of PLG really came from Slack and from their motion, and of their freemium into your really paid motion that they pioneered.
And so I’ll say a couple of things. One is, it is not the perfect motion for every product. I’ll go to Concur. Concur is about expense management and governance. That is not product-led, it is not led by individuals. ‘Cause in order to get the product-led growth, it really has to be something where an individual picks up the product, gets enamored of it, and then it grows. There’s viral growth out from that, from that land and expand motion. And there are great products, and it’s a great motion when you can get it, and it’s a great motion especially for small companies, again, ’cause you don’t need a sales team. Well, you don’t need a heavy sales process to go do that. But it is not applicable for every product and every market. It would be lovely if the world worked that way, but it’s just not. Again, Concur was not product-led growth, will not be product-led growth. It is not Concur’s way of selling. You’re selling into a large organization from the top down.
But again, back to the notion of getting your product out there and iterating, when your product is for an individual — when it is that product-led growth — that flywheel spins so fast. Of the advantages, not only that you don’t have to start with the heavy sales motion, it is also that the feedback is immediate. ‘Cause when they abandon it after 30 seconds or after five minutes or after 10 minutes, you’re done, and that feedback is immediate, as opposed to something that takes a long time to set up and get in there. You still get that feedback, but it’s a longer feedback cycle.
And I just want to touch on one more thing you said. No matter product-led growth or not, as a founder, you are the first seller. Always, always, always, always. You are the person who believes, you’re the one that is out there. Whether it’s a six-month heavy enterprise cycle or the five minutes whatever it is, you are still that first person. You’re the first evangelist, you’re the first salesperson, you’re the first everything regardless of what in motion looks like.
James: That’s right. One thing I’d add to that, I think no question that not every product category is a good fit, if you will, for PLG, but I really, really encourage being thoughtful and finding a way to work your way into an organization through some sort of PLG motion if you can. Salesforce…
By the way, we copied Dropbox. That was sort of the model for us with Power BI. Dropbox was a perfect example where it took you no time to start using it and storing your files. It was easy to share links, other people discovered it. You started to grow an audience, and at some point in time, you tripped over a paywall where you’d stored so much that you had to start paying, and by then, you were sort of hooked. And this notion that you’re delivering value and delivering value and delivering value until you’ve reached a point where you can really ask for revenue, I think, is important.
Salesforce was a company that most of us would believe today as sort of an enterprise sale, but if you look at the way it was originally adopted, if you go back to the early 2000s, it was adopted by individual salespeople as a contact manager initially. It was very, very easy for people to sign up and start using it and storing their contacts. They thought it was cool, and it started to go viral in some ways until it became the solution, if you will, for the organization all up. And so even in this case where you’ve got a business application that ostensibly is sort of an org adoption, finding a way to get individuals at least to raise their hand so that you can then follow up with a land-and-expand motion, I think, is worthy of consideration.
Aseem: Yeah, I think there’s definitely one thing I want to highlight there. And as I meet a lot of companies and founders, a lot of them are focused on the friction-free experience to sign up, but there’s an important element of a friction-free experience to also expand or drive or increase usage.
One of our partners, Karan, who was early and who’s sort of a resident expert in PLG, often talks about this thing of, take the floodgates off the product and let people use and fall in love with the product. And going back to James, what you were saying, which is once you experience the product in its fullness and its glamour and its glitz, and you start using all the features and bells and whistles, then I think it’s a little bit of like… You, as a founder, can then start to be in a position where revenue just becomes second nature. So it’s a super interesting point that you guys both made, and I want the founders and the entrepreneurs, and the early builders to take that away for sure.
They say hindsight is always 20/20, so I did want to ask you that question of what is something you wish you would’ve known if you were still running these companies or divisions? Is there something that we don’t know that would be worthwhile discussing on that piece?
James: I can go first. I’m not that far out, so I’m not sure I’ve learned my lessons yet. But I will say just taking a step back has been really valuable. You always try when you’re in the role to make sure that you’re taking a step back and looking at what the whole world actually looks like. It’s so easy, especially in these big businesses, to get lost in the minutiae. There’s always a crisis, there’s always something on fire, and it’s easy to get lost in that. Consciously trying to take… I tried when I was in the role, but now I’ve been out of the role for a few months, wow. Should have spent more time really looking over the whole landscape.
So the Power BI that the world knows, powerbi.com, is really Power BI V2. There was a Power BI prior to powerbi.com, and it was a set of plugins, actually, for Microsoft Excel. So there was something called Power View, Power Pivot, Power Query. It was sort of the Power family of capabilities that were add-ins to Excel, and you would share these artifacts just like you would share an Excel document. There was no SaaS service. And the learning from that was that, notwithstanding that Microsoft Excel is universally used, the ability to get people to discover add-ins and to have the right version of Excel, and to download and install that into Excel and then to share these artifacts, was really, really awkward. And it wasn’t a success, ultimately.
And it wasn’t until we stepped back and realized that this sort of SaaS model where you can really provide this complete and total five seconds to sign up, five minutes to wow experience, where you don’t put on the user this burden of collecting all the pieces and bringing them together in order to create an experience, was a learning for the team and one that I wish perhaps that the team had learned earlier. But you learn, and you move on.
Aseem: Yeah I mean, goes back to the signup or the usage tax, take it down as much as possible to drive more adoption. We talked about this a little bit as we introduced each one of you, but I’d like to ask you both this question of, what did you think of your biggest competitor at that time? I’ll let you pick who the competition is, but hint, hint, they just might be on this call. What did you think they did well, and what was sort of the thing that kept you up at night? And James, let’s start with you first — if that’s okay.
James: Sure. So look, Tableau was the competitor, full stop. Tableau and Qlik, but really Tableau. Microsoft, for a very long time, had been a leader in enterprise BI. It had some wonderful products, very large customers, and had built a very successful business, but missed this whole self-service BI opportunity that Tableau, and Qlik to perhaps a slightly lesser extent, had driven in the marketplace.
So we were in some ways a little bit fortunate in that we were fast following into a market that clearly had legs. Tableau proved that the world desperately wanted and needed the ability to engage in business intelligence, the ability to analyze data, without being completely dependent on an IT organization to make it possible. This ability to give analysts the power to go connect to data and get value from data was clearly a latent need in the marketplace that had been filled, and that we had missed.
So we thought that they did two things very well, and Mark hit on both of these earlier. Number one, they built an experience that allowed you to very, very quickly, easily, with minimum friction go get the product and start using it and getting value from it without a bunch of handholding, with minimized friction. And two, they built an enormous ecosystem of fans. They built an ecosystem of product experts, they built a community, they built love around the product. And those two things we certainly wanted to mimic, and then we wanted to bring our own unique view on what that market could be and should be.
I always say that we took Tableau and Tableau-ed Tableau on the server side. Tableau had this desktop offering where you could download and begin doing the analytics and create these artifacts, but if you wanted to share those insights, you needed to have a Tableau server set up, and now you’re back to have an IT involved, and having servers and maintaining infrastructure. And so the ability to allow anyone to sign up very quickly and begin sharing those insights, to have an entire organization literally in a matter of minutes have access to the insights through this organized SaaS capability, was where we innovated to try and out-Tableau Tableau, if you will. But we certainly took our cues from all the things that they did right.
Aseem: That’s wonderful. Mark, flipping to you.
Mark: Well, it’s the flip side of the coin. I think James hit on a lot of the exact right points on how we saw the world as well, and how we saw Power BI, which, no surprise, Power BI was the competitor, right? Yes, there was Qlik, and there was MicroStrategy, and there were a few others hanging around, but it is a duopoly of Tableau and Power BI out there in the market.
And I agree with James on what Power BI did really well. We throw all these things in, and we say, “This is business intelligence,” or, “This is analytics,” right? There are variants on this in where the products come together, where there’s overlap. Because Tableau really started as a tool for the analyst to explore data. It was not a dashboard creation. Dashboards weren’t even possible until 10 years into the existence of the product.
And what Power BI did very well was pick up on that and pick it up from that point going forward, it was a very good way to create dashboards and disseminate dashboards, where that was the center of the product, where that was… 10 years into Tableau’s journey the first dashboard was created. It was an ancillary effect, but also where it became viral, where that became something more than was just in the hair and the hands of the analysts.
And I think part of Tableau’s magic is we expanded the definition of analysts hugely, but it still was a minority in the organization. It was dashboards that could be consumed by anyone in the organization that really helped that spread, and that was the sweet spot where Power BI came in. And it became a very, very good dashboard creation tool and dashboard dissemination. And getting the advantage of coming second is you didn’t have to worry about the desktop, didn’t have to worry about being on-prem, it was at that point in time where you could focus on a SaaS service that provided an easier experience for sure, because no one actually likes installing software, shockingly. If you can avoid installing software, it’s just better.
And then the other huge thing that kept us up at night always was Microsoft as a whole, right? It wasn’t just us against Power BI, it was the mammoth thing that was and is Microsoft, right? A go-to-market motion and power that is not just the Microsoft sales team but the biggest reseller network on the planet. The way that it gets in with CIOs and with IT departments, right? I mean, there are Microsoft shops. And then our biggest worry were the deals not where we lost head-to-head, ’cause we did very well there, it was where we weren’t even in the deal ’cause they were just like, “Well we bought an E5 license for Office 365, we have analytics, we’re done.” And there was never a Tableau discussion. That was always a worry when competing with not just Power BI but with Microsoft, because of the reach that Microsoft has.
Aseem: And one of the things that our founders often start to think about, and sometimes pretty early, is this whole notion of GTM and channel. They look at the Microsofts, the behemoths of the world, and they’re like, “Oh, I should just go work with a channel partner,” pretty early. But I think therein sometimes lies the fine balance of — get to true product market fit. Get your first big wins and then get to channel. Versus — starting to think of channel pretty early. Because otherwise you just get spread too thin. And that’s one of the things that we always try to be super cognizant about when we work with teams we invest in and be like, “Hey, what’s the right time for that channel mix to really start lighting up?”
We can go on and on on this topic, but I wanted to shift gears. So much great stuff to talk about. We can’t leave this without talking about gen AI, all that’s happening in the generative AI space with LLMs. How do you folks think that the data analytics category is evolving with these new advances? What should entrepreneurs keep in mind in terms of creating value or building on top of the goodness that gen AI has to offer?
Mark: I’m happy to dive in. I think it’s a really exciting time. Gen AI is the latest great user interface to come on top of this, right? Now finally, there’s a model that really looks like and feels like human language, and can give you a user interface that feels like that because it is… As you work with data, data’s only useful if you can put it into a model that helps you understand the world. That’s it. Data by itself is not useful. One of my favorite sayings is, “Every model’s wrong, some models are useful.” And it is finding those models of the world that are good that help you understand.
And I think what has really been eye-opening with gen AI and the LLMs that are coming out is how good and how powerful that model of the world is, based on language and what… understands. And I think it’s going to have a big effect, and what I’m really excited about is especially the messy part of data, right? Because the messy part of data is not like the output that you see coming out of Tableau. That’s the beautiful, amazing, magical part, that’s the tip of the iceberg. The ugly underbelly is everything that it took to get that data to that point, where that beautiful understanding could come out of.
And I think LLMs have a real chance of helping that process hugely because this has always been something that — well, not always, but now that we’re in this world where we have so much data, machines can do that better than humans. And we need that help ’cause the human intellect, we’re just swimming in these seas of data, and there’s no way that humans are going to sort through all of that. I firmly believe the human intellect will always be the last mile of that, back to that understanding of which models are right and which models are wrong. Where the model is telling you your airplane’s flying 10 feet below the ground, that model’s wrong, and that’s where the human comes in. That’s not going to go away, but the power that you can put at a human intellect’s fingertips is going to increase hugely with these things, because of what’s there.
There is always this trick, though, especially as you get with data and analytics, why it’s fraught with a little bit of danger is that the hallucinations and everything else on what genAI still gets wrong. When you ask for profit, you expect to get profit, and you expect the right answer. This is not a Google search where it’s “Give me four answers, and I’ll pick the right one.” This is data analytics. You’re looking for ground truth, you’re trying to get to it. So it’s going to be fascinating to see how these things, like amazingly powerful, amazing ability to collate so much data and really put sense to it that’s hard for a human mind at that size and scale to do, but then also get to a point where it’s correct enough, how you get the human in the loop, how you really get to answers that you know and believe to be true is going to be really interesting in the next couple of years.
James: I completely agree. I think the thing that we’re going to have to get right, and I think where all the value is going to come from, is getting that human-machine interface right. Where you can arm a human to take advantage of the technology while building confidence in the technology in the human. And the user experience, I think, is where the most work needs to be done.
I think in some ways, as you were talking, Mark, I thought about Trifacta, Paxata, and some of the work that was being done just a couple of years ago where we were trying to classify and do entity extraction and understand the data that you were trying to cleanse, and sort of move into a place where you could get value from it. I think that’s an area where we’ll see huge advances as a result of these large language models, and I’m excited about it because I do think that one of the biggest barriers to getting value from data is cleaning the data and understanding the data and getting it staged to have value extracted from it.
I also think one of the biggest challenges that we’ve had, and one of the holy grails, if you will, is enterprise search. For years and decades, we’ve wanted to unlock the data in the enterprise so that you could ask about things that matter, but the problem’s always been very difficult because the data scale, certainly relative to, say, the internet, is de minimis. And it’s really hard to train models when you’ve got small amounts of data on a relative basis. And I think what we’re seeing now is the ability to take incredibly large volumes of data and make it applicable even to smaller volumes of data, and I think that’s going to potentially unlock the ability to truly, finally, perhaps for the first time get your arms around the small, quote-unquote, data that is your unique data in the enterprise. And I think that’s incredibly exciting.
Aseem: One thing that I keep wondering, James, you talked about plugins, and you talked about integrations into Excel. And I was at Microsoft Build recently, and one of the things I noticed was the plugins are back. And a little bit of these co-pilots are nothing but things sitting on your shoulder advising you what to do, whether it’s changing your settings in the OS or looking at data differently, or enterprise search. And I can’t just wait to see where this goes in terms of, are these still point solutions, are these features, are these platforms? And that’s a space that’s very exciting, just personally, from a productivity standpoint.
You guys both ran massive teams, you built large organizations, but there’s a critical element in every founder’s mind today as they build these companies around hiring talent, hiring great talent, hiring for today, hiring for scale. One question that often hits founders is, when does that shift happen? Should I hire ahead, should I hire for today? What do you folks think is the right way to think about it?
Mark: Delicate judgment. So neither of those answers is completely true. I do love the phrase from Amazon that is, one-way doors, right? Don’t go through any doors that you can’t go back through the other way. Beyond that, of course, you have to build for today to some degree, right? If you’re building for scale before you have scale, that’s a good recipe for never getting to scale, right? You will have so many things to do, you need to worry about. And this is not just organization, this is also for product building, right? Are you going to make non-scalable choices in your product early on? Yes, of course, you are. And you should, because if you don’t, you’re not going to get the feedback, and you’re going to build a really scalable product that nobody wants to use. And so you’re just going to have to feel your way through that on, where am I really at, where am I going, where am I going to be? But focusing on the problems that you have today, don’t worry about, “Well, but three years from now, if it all goes well, I’m going to be here, and I need to build for that.” Again, just don’t paint yourself into any corners, but beyond that, live for today and make choices.
And I always tell people, the biggest ability you have to have in your organization, and it is culture building, which I would say is more important than are you building the exact thing for today or tomorrow? It is building this culture and building a culture, most importantly, that can learn and be empirical about itself. Because, of course, you’re going to make wrong decisions, of course, you’re going to make decisions that don’t scale. That’s okay. You have to make those decisions in order to live today and live to see the next day.
The important part is being able to be empirical about, “Yeah, I sweated blood over that decision a year ago, and now we’ve grown three times, and it’s wrong, and I’ve got to move on.” As long as you can do that and continue to do that and continue to evolve. Because if you’re healthy, again, the decisions you’re making today you hope are wrong three years from now because you hope to be seeing a whole nother size of scale and problem when you get there, as long as you didn’t paint yourself in a corner and as long as you’re very rational and empirical about being able to assess where you’re at and what your problems are and the strengths, and not get married to, “But we grew so fast through here, that must be the right answer.” It was the right answer then, and that’s awesome. The right answer for today is probably different if your business is healthy and growing and changing.
James: Yeah, the one thing I would add very specifically is to be careful about bringing on a big go-to-market machine too early. It’s amazing how many organizations that I’ve seen, and I talk to a lot of founders and even just in the last two weeks, had a conversation with two companies that I think got out in front of their skis by hiring big, expensive enterprise sales leaders before the product was really ready for it. And that can get you upside down.
Salespeople are going to go out, and they’re going to sell. They may sell you into an environment where the product’s not appropriate, where now you start getting all these requirements that are specific to that customer. You can get very distracted, you can start to love to see those big revenue numbers coming in, and you can move from this world where you’re building a scalable machine, where you’re building a product that can drive usage that you can then harvest. You’re sort of going upside down and selling something that you then need to drive usage behind. And I think that’s one of the bigger mistakes that I see repeatedly.
And so in that regard, I would be very careful about, quote-unquote, building for scale too early because it could kill your products and ultimately require a reset later.
Aseem: I think very wise advice. One thing I’ve also observed in meeting these companies is you tend to focus more on top-line versus usage and stickiness, and I think this goes back to our land-and-expand conversation around what is it that’s driving expansion, what is it that’s driving fan creation and fandom of the product? Something for folks to definitely keep in mind. Any particular spaces and companies you folks are excited about?
Mark: I’m excited about what genAI and, more specifically, what LLMs can really do for data, and especially that data, the mussy machinations. Numbers Station, a Madrona-funded company — super excited about what they’re doing, trying to bring that technology here. And I think we’re going to see a whole bunch more innovation in this space on all the messy pipes. What does ETL look like? What does data prep look like? ‘Cause this is the biggest problem. Every customer I talked to at Tableau, like yes, there were things about the active analytics that were still to be done, but that wasn’t their biggest problem by far. It was, how do I find data? How do I understand what that data is? How do I clean it and get it into a shape where it can actually be useful? And this is the problem that companies like Numbers Station are starting to tackle, and I’m really excited about what that can do to really accelerate the usage of data.
James: Yes, and I think that gen AI in general, or LLMs, and the applicability to the business process layer, what applications are going to be enabled, both horizontally and industry-specific operational applications?
We’ve talked a lot today about product-led growth, knowing your customer, understanding your user, learning from them, turning, burning, improving the product. Companies like Interpret and Viable, I think, are examples of interesting applications that are very specifically about gathering as much unstructured information as you can from your users, from your customers, from the market and interpreting it, understanding what you can learn in a way that previously was pretty high touch. One of the things that we did, and one of the things that I found frankly, back to Power BI, we were maniacal about collecting user feedback, and it was both structured and unstructured. We’d do NPS prompts, but we always had a place where people could type in their thoughts, and the ability to analyze that and to understand the trends was always time-consuming. And LLMs are offering an opportunity, I think, in a way that we couldn’t before, automate some of that so that you really can get the learnings very, very quickly and perhaps across far more channels than were previously available.
So, excited about that and all the other potential use cases that we see in business applications.
Aseem: This has been so amazing. Any parting advice for those building companies, founding companies? If there’s one thing you want to leave them with, what would that be?
Mark: I always say, and told my groups all the time, we only get to come to work at the pleasure of our customers. And it’s all about what problem you’re solving for your customer. And if you make happy customers, you will do just fine on the rest of it. Like James said, don’t focus on your go-to-market motion, don’t worry about what your customer… The first thing you got to do is build a product that’s solving a problem for a customer. If you’re solving problems for customers, the rest of it’s still going to be hard work, but it will take care of itself.
James: I would just hammer that home. We had this thing that we coined called Loose LUSRP, logos, usage, satisfaction, revenue, profitability in that order, and sort of walking down that stack. And by logos, I mean the ability to have a brand new customer who’s not a customer today or a user find the product and begin using it. And then you sort of work your way down to, “Okay, now let’s get more usage out of that. Let’s drive satisfaction.” And eventually, you can worry about revenue and perhaps gross margins at the end. Focusing on the user, driving usage, minimizing friction, increasing the velocity of improvements, listening, learning, and improving ultimately is, I think, the key to building a great company and certainly a great product.
Aseem: I want to bring back an example that I heard while I was at Microsoft, was this whole notion of, “We’re in the business of delivering happy meals. Happy meals make happy customers, and happy customers give you more happy customers.” So that was sort of what came to my mind as you guys were both talking about this.
But this has been amazing. Nothing short of fun-filled and a lot of insights that comes from years of you guys’ experience that our listeners and founders will enjoy. Thank you so much, this has been pleasure for me to host this and have this conversation. So on behalf of Madrona and the entire crew, thanks, and more to come.
James: Thanks for having us, Aseem.
Mark: Yeah, thank you, Aseem.
Coral: Thank you for tuning into this episode of Founded & Funded. Be sure to follow us wherever you get your podcasts, and tune in in a couple of weeks for our next episode of Founded and Funded, which featurs Bob Muglia, who just released his first book, The Datapreneurs.
Related Insights
