Ben Horowitz 与 Ali Ghodsi：如何经营一家千亿美元企业

I was like, maybe they’re right. Maybe we should just sell. I remember having that conversation with Ben. He said, “Hey, you can do whatever you want. You can sell. You’re going to make a lot of money and you’ll be super successful in life, but, you know, if you’re like me, you’re going to look back the rest of your life thinking, ‘I missed that one shot. That was the one thing I should have taken all the way. Now I’ll never know how far I could have taken it.’” Could have been.

So, do you want to live with that, or do you want to just have the money? I’ll support whatever you want to do. I really couldn’t care less.

I’m excited to bring back Boss Talk. This was a series that you guys did a few years ago on Clubhouse that was a big hit.

Yeah, we had fun. It was Ben’s idea.

Yeah. Excited to bring it back.

In the spirit of Boss Talk, let’s talk about the first time that you became a boss, in terms of running Databricks. Let’s talk about the moment in 2016 when Databricks wasn’t as smooth as perhaps it should have been and we were looking for a new CEO. Ben, you recommended Ali.

First of all, kudos to Ion Stoica for building the company originally, and to Ben for investing in us and believing in us. I also kind of couldn’t have done the CEO job. Ben basically babysat me the first couple of years—a short, short baby.

I did know what was kind of wrong with the company because I had been there for 2 or 3 years, and I had seen from the inside what we should change and what the issues were. We had an open-source project that actually became very successful thanks to those first 2 or 3 years.

Apache Spark became a worldwide sensation, and we could pride ourselves on the number of downloads of the software. Well, and the Spark Summit.

Yeah, the Spark Summit.

Now the Data + AI Summit.

Yeah, but the problem was that, as is often the case with open source, everyone was just downloading the open-source version. Actually, your biggest enemy is your open-source project. The main thing you have to fight in the market is, “Hey, why can’t I just download the open-source version? Amazon is offering it. The cloud vendors are just offering it. I’m just going to use that.”

That was the biggest challenge that Databricks had at the time, and we needed to make very serious, aggressive pivots internally, which were going to be very painful for a lot of people—for the whole ethos of the company, internally. I had known that for almost a year, so when I got the shot, that’s what we started doing.

The strategy was, “Make Spark the biggest open-source thing.” I can remember it on all the slides. Then Databricks would have the best Spark, but Databricks never necessarily did a lot to make it the best Spark, or didn’t differentiate it enough. That was kind of the first thing Ali did on the product side. Then he hired Ron Gabrisko, which was transformational because that kind of dragged the company into the world.

So, obviously, that was the right decision and paid off. Maybe zooming out, Ben, you’ve worked with all the great CEOs of our time. Where does Ali stand out? What are his superpowers as a CEO and as a boss that have helped contribute to the impact?

Ali’s really good. I always rate CEOs by asking, “If I were running that company, would I do a better job or a worse job?” With Databricks, I’d do a way worse job. He’s good on many, many dimensions.

First of all, he is a real technologist—not a pseudo-technologist like his competitors. I’m sorry. He really knows the product, and he understands the product strategy in detail. He also ran engineering before he was CEO. Mostly, what I worked with him on in the early days was go-to-market and BD, and he’s really good at both of those. That’s where we had to catch up. Snowflake had an amazing go-to-market, and then we needed to deal with big partners.

Every time I got him a little BD tutor, John O’Farrell, who did a nice job, came in and taught Ali how to structure a deal and how to do things. But he learned everything so fast. Probably the thing that he does that I wish I could get all our CEOs to do is that he doesn’t hesitate. He trusts his eye. He’ll see something, and he doesn’t know if it’s right, but he’ll investigate it.

If you look at the strategy changes Databricks has had, one big one was building a data warehouse. That was a pretty big swing and a seemingly quixotic, insane idea given where they were. But he was both paranoid enough to know that it could be an issue and confident enough in himself to go deep enough to decide whether to do it or not, as opposed to ignoring it and thinking, “These guys are trying to kill me. I don’t want to see it.” That’s what a lot of CEOs do. There are a lot of elements to that job. It’s a very complicated job.

Ali, talk more about the journey—about evolving from an academic and a technologist to someone commercial. It’s a journey our CEOs go through. Talk about what it was like for you, in the context of what others can learn from it.

We were in academia, so we were scientists. Then I led engineering and product, so I had to learn how to build a product and get product-market fit. Then I became CEO. Each of these has different challenges. I think the thing that is common to all of them is that you really have to understand and be extremely good at the task at hand.

Number 1, admit that you don’t actually know everything about the job. The first step of Alcoholics Anonymous is, “Admit you have a problem.” Number 2, be a student and learn everything you can about it. Go all the way down to the details, try to learn from the best, and work your butt off. You know nothing. You’re zero, right? You know nothing about writing reliable software.

For me, that was the same thing I tried to learn. I tried to network with the best heads of engineering and the best heads of product. I tried to read every book I could. I got as much as I could out of Ben and Marc. I read all of their blogs, all of their books, and everybody else’s.

Then you do research, and you start looking for the number-one product manager by reputation in the market right now. Can you get 30 minutes with that person? Just sit down. They’re not going to join you because your company is too crappy and too small, but can you get 30 minutes with them? Can you get a dinner with them? Can you get a breakfast with them? Then ask them lots and lots and lots of dumb questions, and they’ll tell you. They’ll happily just tell you, “Here’s how I run. Here’s how I do it. The other guys are wrong.”

They’ll give you a playbook, and you can go compare it. You can go to the next person and say, “Hey, this is the playbook I heard from the last person.” They’ll say, “No, no, that’s totally wrong. You don’t do it that way. Here’s how I do it.” Very soon, you learn enough. If you really have grit and work hard, you’re going to be able to do great things.

That’s about you yourself. But also, if you hire a great team, as a leader you alone can’t do much. Can you hire the best people out there? That’s also part of it. Do you know what great looks like? Have you interviewed all the best people? Can you now sell them and get the best people to come work for you?

Once you start assembling a team of excellent people, they will uplift you. This is the managerial leverage that I learned from Ben, which is from High Output Management by Andy Grove. Are they so great that you’re learning from them? I was a great head of engineering because the people who worked underneath me were doing amazing things. I was just standing on their shoulders.

You have to instill that in everybody else, recursively, so that you end up with an amazing, killer team, and you have to continue doing that. For engineering, it wasn’t actually that hard because I had written a lot of software. But now you’re CEO, so you have to do that for marketing and sales, where you’re really clueless and probably all of your instincts are wrong and your intuition is completely wrong.

They were super helpful because they had done it with Loudcloud and Opsware, so they knew how to build a B2B machine and how the game was played. But you have to do it again, and now you’re doing it in a field where you’re really clueless. Can you be clairvoyant and see the truth, or do you want to lie to yourself? That’s where a lot of founders make mistakes. They’ll do well in their own archetype, but when they have to step outside of their own archetype, they make a mistake. They hire people who are like their own archetype in other roles, where that could be lethal.

By the way, that's how we started Databricks: I think everybody who was running anything had a PhD in computer science.

Yeah, including sales.

Yeah. That's probably the number-one mistake. You go, “Okay, well, I'm an engineer, so I want a sales guy who can talk to me and understands engineering.” That's not really a good criterion for sales.

One thing that I'm good at is, rather than telling somebody that they're stupid and hurting their feelings, I'll ask them a really fucked-up question. I did it in a board meeting. I said, “Could you help me with the math on this? I don't understand the math.”

Actually, it was worse. I said, “Help me with the—I'm just trying to understand basic math. You have all these numbers on the slide.”

And if you said that your conversion ratio is 5%, I'd say, “But I can't divide either of those two numbers to get 5%.” Then the person freaked out and said, “No, no, don't freak out. Just tell me which of the two numbers I divide to get 5%, because I've divided all of them and none of them is 5.”

“Am I going to be fired?” He does a much better version of that. If somebody's really screwed something up or is messing up, he'll go, “How do you think it's going?”

And I was like, since he told me that, “Oh, yeah, that's a better way to do it. That's even better.”

So, yeah. He's a very good student.

Can I reframe that? There's this book called Radical Candor, and I think people take it too far and misunderstand it. But I think the essence of that book is that, if the feedback is, “Are you criticizing me?”

Yeah.

“Are you saying I'm stupid? I can't do the division, because my point is not about the 5%. I was trying to make a different point, and now you're just—this is a cheap shot, and now I'm hurt.” And by the way, I think you're wrong. It's not 5%; I said 6.5%.

So are you criticizing me, or is it, “No, no, I'm here to help you”? I can, like, not help you, but if you beg me for help, maybe I'll help you. So which of the 2 modes?

If you can get people into the mode of, “Oh, wow, I'm being helped. They're helping me, and I'm going to get further ahead in my career and be more successful”—“Please, no, no, please don't leave. Come back and tell me more, because I'm taking notes here”—then you can flip to that mode.

A lot of feedback can be recast into, “I'm just here to help you, but feel free to completely ignore this advice. But if you want to be really successful, if you want to get that job or that project next time, if you did it this way, you probably would have had a higher probability of getting that. But I don't care. You do whatever you want.” People are much more receptive. They're like, “No, no, no, please. I want to know more.”

Yeah. Well, I think the frequency of it helps a lot, too. If I see you once a year at your review and tell you what's wrong with you, you're going to be offended. No matter what it is, no matter how wrong it is, no matter how correct I am, it's going to be offensive.

But if every day I see you doing something I don't like, I go, “No, don't do it that way. Do it this way,” then you get desensitized to it. I think the mistake a lot of engineers, particularly engineers, make is that they just don't say what they think when they think it, because they're afraid of hurting someone's feelings.

That's how you save their feelings, because they're used to you. You're always doing that, and you're doing it with everybody. They see it. They're like, “Oh, yeah, fucking Ben's an asshole. He's always doing this, but that's how he is, and that's how we work, and it's no problem.”

As opposed to the hammer, where you try to put it in a fucked-up sandwich: “Oh, you do this really well, but this is all fucked up, and this is good.” People are like, “Now, in my written review, you're telling me for the first time that this is all fucked up. Fuck you.”

This is very common, and you can see this in the industry. The extreme version of it is they get fired, right? Then the head of HR talks to them, and they're like, “Did you see this coming? It was obvious, right? You knew this.” “No, I had no idea.” “Wait, you didn't get any feedback on this?” “No, I only got thumbs-up all along for a whole year, so I'm in shock.” This is super common, right?

So, maybe on the topic of managing talent, you have this incredibly high-intensity culture at Databricks. There was this thread recently in our CEO thread where they asked everyone, but you had a great response: “Hey, we have 50 people. How do we scale? We have this culture of 996, right? You work 9 to 9, 6 days a week. How have you scaled that intensity all the way to 10,000 employees?”

I think you start with setting the tone at the top. If you're the hardest-working person, everything will kind of take care of itself from there on. If you're not working hard, it's very hard. If you have a double standard—I mean, Ben has a whole book about that, What You Do Is Who You Are—what you do is who you are.

If you're working extremely, extremely hard, the rest of the organization will as well. Are you calling people at 9 p.m., 10 p.m.? Are you working weekends? Do they expect you to—not that you expect them to, and you're going to be angry and yell at them if they're not dropping everything for you, but the fact that they just know that Ali is working 24/7, 7 days a week, and that he's working at 11 p.m. or 2 a.m. or whatever it is? I think that gets a lot of it done.

You can vet for this when you hire people.

It's got to be careful, because the people who say they're going to work the hardest are not the ones who work the hardest.

It's the opposite.

Yeah, 100% true, right? The best way to vet for this is to do backdoor references. If I ask someone, “Hey, how was Sarah? Did you like her? Was she great?” they're always going to say, “Yeah, she was great,” right?

But they're going to be much more honest if you ask them, “How much does she grind the midnight oil? Is she—” They'll tell you right away. It's like, “Oh my God, she works like crazy.” Or, “I think she has a good balance.” You can suss that out very easily from backdoor references. People will remember those people, and they'll just offer it up and say, “Oh, that person was nuts. They were working 24/7.”

I think that way you can get people who are hardworking. By the way, I don't want to overemphasize it. I don't think everything is just “work harder.” You have to also work smarter, and you want to make sure that it's sustainable.

I can work insanely hard. I'm motivated. Everybody has a different threshold for how hard they can work. I don't think you want a culture where people are burning out. You really should avoid that. In fact, at Databricks, I'm very often going in and saying, “Hey, this team's scores are really bad on work-life balance. What are you doing about it?” Or, “You guys should take several days off. You should do some offsites or do something.”

We actually go in if we see that there are some groups—and other groups at Databricks, their work-life-balance scores are like 100%. They're slacking off. So then it's kind of the opposite. But I do think that you can make up for that.

I think that also means setting the expectation. One of my competitors, Frank Slootman, wrote a book called Amp It Up. It's a great book on how you get execution into a company—how you get a high-performance culture where everybody's always trying to excel and do better and better. That's a good book if you want to study how he's doing it at scale in bigger companies. It's highly recommended reading as well.

Yeah. And I think a lot of it at his scale ends up being things like organizational design. Do people feel like they're having an impact? If people feel like they're having an impact and they're good, then they'll work very hard.

But if you're in some kind of weird 3-legged race that the CEO has constructed, where everybody's got dependencies on everybody else, it just doesn't matter. You'll have a lot of people go, “I know if I work hard, it's not going to make a difference. Why would I do that?” You can't overcome that with rah-rah, leading by example, or anything else. That's just fundamental to how it is.

In any company of any scale, even at our scale, there are some groups who can have an impact and work extremely hard, and then groups who have less impact will work less hard. You just see that.

People who are motivated and feel excited about work, but don't see the impact that they're having, are going to work way, way, way harder than if you're demoralized and feel like it's not going well. If you're not having an impact and don't have any autonomy, you're just not going to want to. You're kind of depressed, sitting down and working.

I do think there's one thing here where leaders can really help, which is to make your team feel like they're winning and that they're doing a great job. You can ask more from people, but if I feel like, hey, I'm losing and everything we're doing is wrong, and I'm putting in all these hours and it's stupid—what's the point of this?—then people don't want to work. So I think it's about feeling like we're winning, like we're the winning team. We're winning, and wow, they're expecting more from me. Then I think you can get the motivation you need in people.

Yeah. Yeah. Which is why, by the way, the hard job is when you aren't winning.

Yeah, to get the output, particularly in Silicon Valley, because you're battling attrition and this and that. To get things on the right track, that takes a whole different kind of level of technique and storytelling, and showing you how you could be winning and all that kind of stuff. That gets very, very complicated.

We've both done that, right?

Yeah. There's been phases in our companies' lives where we weren't winning. I mean, especially the story you had in The Hard Thing About Hard Things, which is probably the best business book I've read. I read it, by the way, before starting Databricks, and it influenced us a lot. Those are super important.

Yeah, that's the difficult part. That's such an important point, because even if you're winning, people have got to feel like they're winning. But if you're not winning, getting them to feel like you're winning means saying, "We have a path to winning."

Yeah, we have a path.

We have a path, and it's rock-solid. It's going to work, but it demands sacrifice from all of us. There is no feeling as good as when you're not winning and then you get to winning.

That's the best feeling. You can't replicate that once you're super successful. You never can quite get that feeling again.

Yeah, that's true. But you also never feel that horrible pain again.

Well, it's easier to be the underdog in some ways, right? You have nothing to lose in some ways.

In most ways, not.

Well, I want to explore this leading from the top, because that was kind of the first thing you started with. We actually hired an ex-Databricks employee at a16z, so we have some inside scoop on your leadership style. One of the things he said was—and Ben sort of touched on this too—but you have this amazing ability to be strategic and help your team focus, but you're also very much in the weeds. You're giving product feedback, you respond to emails super quickly, and to product launch emails, no matter how small they are, you'll respond, "Congrats," which he found hugely motivating. How do you do all that? Where do you fly high, and where do you fly low?

By the way, I respond even to progress reports on all those products, and I follow them in detail, every one of them. I try to respond to every product.

Insane.

I respond. But look, I think if you're just going to fly high and give high-level inspirational speeches and then say, "We'll trust and delegate to people," it's not going to work. My way is that you've got to get in the weeds. You've got to understand this. This is back to what I said at the very beginning: How do you become great at being head of engineering? How do you hire a great head of marketing? The only way you can do that is by being really excellent at it. You need to study the game and become the best, so I try to stay tuned to all of these things.

There's this quote: "If you do everything, you will win." The question is: Have you done everything?

Exactly. Exactly. Exactly. So, yeah, it takes a lot of effort. You need to learn all your keyboard shortcuts.

We used to have one of the culture principles: "Be a co-founder." We didn't want to have any employees at Databricks; we just wanted co-founders. The key point was, hey, you're kind of the owner of this company. You're not just a renter. Come here, and, yeah, we can talk about it. You can suggest an idea. You might have just joined and be straight out of school, and you might have a great idea for a product. Tell me about it. I'm happy to push it.

It's making people feel like they have an impact and they're inspired. Back to Ben's point, then it's going to be much more exciting for them than following some bureaucracy. I don't follow the bureaucracy, basically. I go talk to anyone I like. I try to go to the person who is actually closest to the work being done at any given time.

But there are some tricks and rules around how you do that without breaking the whole organization. You can't just willy-nilly talk to anyone. But, yeah, that's part of it.

Yeah, listening and giving direction is very different. If you give direction, you can cause a lot of chaos. But if you go talk to people, listen to understand the problem, and then send it back down the chain of command, that tends to work very, very well.

Generally, if you're a CEO and you don't fly low and fast, it's going to be a mess, because you never get the truth. The truth never makes it to you through your people. If I go talk to Ali's executive staff about what's going on in their organization, or anybody's, first of all, they're going to spin it. Second of all, they don't actually know.

You need to help them debug their organizations, because they've got a million things going on. They're also kind of going to the problem, going to the bottleneck, trying to figure out what's happening. It's just a very unreliable source of information. All the knowledge in a company is with the individual contributors who are doing the work and with the customers. There's no knowledge with the people who are talking to you as CEO, who are on your staff. That's not the way information moves.

Ali is super fast, which enables him to go super low. But at any given time, the way to think about it as a CEO is that you're not spending the exact amount of attention on HR as you are on the key engineering project, or on the key sales competitive deals. You don't address everything evenly. You can never do that. It's just a bad idea.

You'll probably get to everything eventually, but you're not spending the same amount of time on every single department. The org chart is not the way the company works.

It's just a communication architecture.

Yeah. I think the best way I would say it is that it's kind of like a T. You want to be broad, and then you have the leg that goes down and goes really, really deep. You want to do that anchoring, and the key thing is to have a really good priority order of what's most important and drop everything else.

You drop that T and go really, really low. It might be HR. I might be deep-diving all the way down to HR, looking at our HR handbook, our policy, everything: Who is this person? What happened? Why is this happening in that group? What's going on in that group? What's the culture in that group? What happened here?

You might want to do that. It might be existential for your company, as we've seen some companies go under because of HR problems or ethical issues that were going on. So I think having a really good priority order is really important. Some executives just want to have their ducks in a row: I have my weekly 1-on-1s, I have my weekly staff meeting, I have my weekly this, and then I do this, and then we follow the rules and do all of this. That's just the top part of the T, and then there's nothing that goes deep. That's the issue, I think.

Yeah. Over-systematizing or making it symmetrical—you don't have to have 1-on-1s with all your staff at the same frequency. Some of them you can meet with very seldom, while for others everything is different. Every part of the company is different. You may need to meet with somebody every day.

Yeah.

Other people you can meet once a quarter for now because it's just not that serious. You can't get caught up in making everything fair and symmetric. Particularly with your staff, they've got to be able to deal.

This is actually the biggest conversation that I had with Ali early on: If they can't do it, they can't do it.

That's it. It's a wrap.

Yeah. Yeah. Yeah. Don't try to fix them. They can't be fixed. It's not going to happen.

And, you know, it's a sad lesson, but an important lesson. I actually want to turn the conversation to an area that Ben was saying you had to catch up on, at least in the beginning, which is the BD dealmaking stuff. That's interesting to me just because I think of you as a consummate dealmaker now. I feel like you're playing chess while everyone else is playing checkers.

I want to go back to 2017, with maybe one of the first game-changing deals that you guys did, and that was the deal with Microsoft. Can you guys talk a little bit more about how that deal came about? Is there anything you would do differently? By the way, founders still to this day ask us about it because it's sort of a model for how they'd like to do deals.

Yeah, maybe I should start by saying that we had tried to get close to Microsoft for a long while. I think Ben had told us, “You need to—that’s an important partner because they have the biggest distribution channel. They have 60,000 sellers today. If you can unlock that in any small way, it’s going to be a game changer for you.”

I had been CEO for a year, so I’d been trying hard to get in there. Many people offered me, “Hey, I actually know Satya, so I’m going to get you introduced.” I got multiple introductions to Satya. He either never responded or just CCed his EA, and it went to the EA: “We’re still trying to find time. He’s been so busy this last 6 months.”

Then he had a meeting with Ben—I think he was actually here at a16z—and they just talked. I wasn’t actually in the loop, and then he called me up and said, “Hey, I talked to Satya, and I think he’s excited. He wants to do this.”

I saved the email. Ben introduced me to Satya, and this was, I think, 3 or 4 a.m. I was in New York. The email went to Satya, and then Satya added 4 or 5 people to the email thread, and then they added 4 or 5 people. Within an hour, I had 25 emails in my inbox.

Suddenly, all these people who had not been responding to my emails from Microsoft—right after Satya CCed the next person—they were all saying, “Hey, I’m clearing my calendar. I’d love to meet you. Do you have any time in the next 2 or 3 days?” But really, the original pitch of what the give and get was came from Ben and Satya at a16z. They figured it out, and I wasn’t actually even there.

So we had some luck, and then Ali did quite a few things that were very, very effective. The luck was that, at the time, with deals with big companies, there was always a timing element. There was a company called Hortonworks that had a deal with Microsoft to provide some similar kind of functionality, and they were basically putting a gun to Microsoft’s head, saying, “You pay us more money, or we’re going to pull our product.”

They were on-premises and in the cloud, so it was a big mismatch, too.

Microsoft was super pissed at them and wanted to stick it to them. You had Satya going, “I think this company’s interesting,” and then this ground-level thing going, “We want to [expletive] these guys.” That opened enough of a door to get it going.

One of the most important things in the deal—and John O’Farrell really emphasized this for both of us—was that you’ve got to get them to put enough in. They’re such a big company that they’re going to lose interest many times. If you don’t have them write you such a big check that somebody in there is going to get fired if it doesn’t go well, it doesn’t matter if you get the deal—you’re going to lose the deal.

What we did was say, “Okay, give us a forecast. We’re a little company. We can’t afford to do this deal. We can only afford to have 1 partner, so give us a forecast of what you’ll do.”

Yeah, because our engineers are busy. They’re going to do this integration that wipes out 12 months of our roadmap. We don’t have anything else. You guys have many thousands of engineers, so this is the only one we can do.

Yeah, so we said, “We think you can sell the most, but we don’t know. What’s your forecast?” We challenged their manhood a little bit. They came out with this big-ass forecast, and we were like, “Okay, great. Just give us a little portion of that.”

It was a huge deal. It was a lot of it.

Yeah, and then Ali said, “Look, when we got all the way down to the deal, if I don’t get this number, Ben’s going to fire me.”

And so, can you help me out?

It was very interesting. Bad cop.

It was a very interesting dynamic. John O’Farrell had strategized with us and told us that they have to do a big pre-commit because then they have skin in the game. Otherwise, they’re just going to forget. They’ll do the PR, but then they’ll forget about you.

When we were trying to get that from Microsoft, I remember I was talking to Takeshi Numoto, who is one of the main brains at Microsoft, one of the key strategists there. His thing was, “I don’t want to give you a big commit because you’re such a small company. I’m worried you’ll take this money, get drunk off it, and not do anything afterward.”

I had to really convince him: “No, I’m extremely hungry. There’s no way I will continue to have crazy appetites. Don’t worry about it.” So both sides were worried about different things.

But the give and get you mentioned in the beginning was important: they had a gap in the product portfolio, right? They were competing with AWS. They had a gap at the time, and we had a great product. They had an amazing distribution channel. In these BD deals, there always has to be a give and get that’s commensurate. This is why most of these deals fall apart and don’t work.

There has to be something that you, as a small player, can give that they don’t have. Usually, you don’t have anything to give them. Usually, I find all these small companies show up and come, for instance, to Databricks now and say, “We’d love for you to partner with us.” But what am I getting out of it, right? You don’t report to me; I don’t report to you.

The moment we’ve closed the deal, if it’s not good for me, neither of us will just do our side of the bargain. So there has to be something in the deal dynamics, in the construct, that’s inherently extremely beneficial to both sides. There has to be a trade that makes sense. Microsoft really wanted that product; we really wanted their distribution channel.

The other thing that I think a lot of entrepreneurs don’t understand is that, in any big deal of that size, you lose at least 3 times before you win it. We lost that deal—

10 times.

10 times. Including the day before we were supposed to launch it. The antibodies came out of the company, and Ali had to fly up to Redmond and sit there.

There was 1 engineer who just said, “Not doing this. This is not going to go. We don’t...” They actually put a guy in place at Microsoft who was super—he had a great reputation, but he was a builder, so he had huge problems with this. He was like, “This is not a product I built. Why would I make this successful?”

Usually, there are many times when these deals can die. If you don’t have grit, those deals will die, because this deal died multiple times. It was completely over. It was completely blocked by some executive who said, “Absolutely not. I’m blocking it. It’s vetoed. It’s over.” No one wanted to overrule him.

You have to go in there and work. The only way we did it—I like to call it the Nerd Bird. I would take the San Francisco–Seattle flight up there. I was up there so much that I knew all the buildings, all the rooms, everything. You just have to spend time on the ground, talk to as many people as possible, and sort of influence that organization from within.

With all the difficulty of the deal, and Microsoft being Microsoft, they’ve been as good a partner as not only we’ve had at Databricks, but in the entire portfolio. They’ve really lived up to and delivered what they said they would do, which I think you have to give huge credit for. In the whole Gates and Ballmer era, they were never that good a partner to anybody, and Satya’s really turned that around. They’ve been fantastic with us.

This was around the time Satya had taken over, and he was giving everyone at Microsoft the book Mindset, which is about growth mindset. There was this aura in the air that we should try. Let’s try to make things happen. Let’s have a growth mindset here. Let’s see if there’s a way we can partner.

This would have been impossible 5 years earlier, so kudos to Satya. They put us on the map, and he’s been a great partner ever since. Whenever there have been issues, they’ve always resolved them. We are very thankful. We wouldn’t be where we are without them.

Yeah, just amazing. Really amazing. I want to open up the conversation to dealmaking more broadly. Now that you’re not a small company anymore and you’re a big company making acquisitions—Tabular, Neon, MosaicML, just to name a few—what is your approach in terms of when to build versus when to buy? How do you think about acquisitions more broadly?

Yeah, I mean, what we try not to do—so let’s start with a simple thing—is buy revenue. A lot of companies, especially at scale, will buy revenue. They’ll look at a company and say, “Hey, this company is this size. We’ll just buy that company, put more salespeople on it, and then we can accelerate the revenue we’re buying.” That’s how they’re doing it. We’re not doing that.

What we’re really doing is, number 1, spending a lot of time with the team and the founders. We’re trying to see, “Hey, can we build together? You come here and you build together.” That’s very different from that buying-revenue model.

The buying-revenue model oftentimes means you part ways with the CEO from day 1. You can see the big companies; they literally have a plan. I have some executives who come from these big companies, and they say, “Our plan usually is to part ways with the CEO. You make a deal, and the CEO can leave.”

But also, the key people in those companies quickly leave—all of them, the top management. Then you keep promoting the people from below who couldn’t get promoted before, and eventually you bring in your own people to take over the company. Then the company is dead.

There’s nothing left of it, and there’s no integration between that asset that you bought and the platform that you have. To avoid all of those problems, can you get people who really feel like they’re your co-founders?

We spent an enormous amount of time with the company we were buying: Who are the founders? How do they work? Are we culturally the same? We spent time with them. Do we get along? Do we see the world the same way? Are we going to click? Are we going to do this together? Are we going to be able to build together over the next 5 years? That’s where we spend our time, number 1.

Number 2, we spend a lot of time on the product. What’s the product experience? How would we integrate this? What would it look like? How much can we rewrite? Can we not rewrite it? What programming language would you write it in? People are like, “Why? That’s such a dumb question. Why do you want to do that? What does that matter?” No, because we’re going to integrate the codebases, right? The build systems won’t work. It’s not going to even compile.

We spend a huge amount of time on the product, talking to customers, understanding what the excitement around that product looks like, and how the integration would look. The last thing we do is look at the financials: What’s the revenue multiple? How much can we grow it? What’s the 3-year plan, 5-year plan, and so on?

I feel like big companies’ corporate development departments do it exactly in the reverse order. They start with, “Hey, the revenue is this, but we could accelerate it, and the multiple is so low. In my Excel sheet here, this makes perfect sense.” Then, second, they ask, “Is this a good product?” And lastly, it’s, “How do we convince these knuckleheads? We probably don’t want to have them here, but we’ve got to pay them off somehow.”

I think if you think about it that way, you get more longevity out of it.

Yeah. And this is really the thing that people get wrong on the go-to-market side. If you’ve got multiple product architectures, that’s going to mean multiple sales engineering forces, multiple post-sales things, and your entire sales efficiency is going to go through the floor.

Because they have a keen eye on that, everything they buy ends up looking like a Databricks product. That work is going in. They’re not just selling some shit to get some money to go on a corporate development thing. I would say so many times, when you bring in a professional CEO, this is what they screw up because they don’t understand that.

Engineering goes, “Yeah, yeah, we can take it on. There’s another set of engineers. We don’t care if they work on that.” Engineering gets less efficient too, but it wrecks the field. And then the customers hate it because they’re like, “Okay, I’ve got to learn another access control model. I’ve got to do this.” These are not things anybody wants to be part of.

Yeah, 100%. This is the go-to-market side that you’re worried about: the experience those customers will have. They’re going to come back immediately and say, “Hey, we were already upset about these things before the acquisition. Maybe you can fix them now.”

It’s like, no, actually, several of those people quit, and now we’re just going to work on integration. That thing just got pushed out another 2 years. You don’t want to be in that situation.

There are a lot of companies that do that, and, by the way, what they’re doing works revenue-wise. They’re getting the revenue, and the stock swap works. If the multiple is right, it’s a creative deal temporarily.

Yeah, it works. In the 1st year, you get the bump in revenue, and you get a 2nd-year boost in revenue growth as well. The financial engineering actually works great for those companies. It’s just that, long term, it ends up being a bag of crap that doesn’t work together.

And it affects a brand.

One of the reasons Databricks is so powerful is that all their customers want to buy all their products because they’re like, “We know that’s the best software we buy.” As soon as you start chipping away at that with these financial strategies, you can’t get it back because the reputation is every customer’s experience. There’s no marketing through that.

It’s the best software because it was written by the engineers and built by those who were the best, including the acquisitions that we got. They were phenomenal people who came in, and since we gelled, they continued building it.

That’s why it’s great. Back to the question of who you’re getting into your company.

Yeah. That’s the other thing, right? You can buy something that’s got a lot of sales, but you’re downgrading your whole company. Ross Perot actually wrote about this in Citizen Perot. His biggest fear, which definitely came true, was that he built this elite thing at EDS, and then they would acquire IT departments. He was like, “They’re going to absorb us, not vice versa.” And that does happen.

Mmm-hmm.

There is 1 really successful company that we never acquired. I always vetoed it whenever it came up because I just thought the quality of their employee base was not great, and I didn’t want it to dilute Databricks. From every other angle, that deal always made sense, and I always vetoed it because I felt that they were all going to quit or be super unhappy. I thought, “Let’s just not do it.”

It’s also why mergers of equals are hard, because the cultures aren’t equal. The people aren’t equal. What made you feel that way? You just spent time with them, and they didn’t exude the Databricks culture?

Well, look, it’s like with everything else. It’s like when we were grading students at the university. The rock stars are super easy to find. They’re right there. The people who are really, really bad aren’t hard to identify either. Then there are people in the middle, and that’s the gray zone.

This was a company where I felt the talent was not phenomenal, and you don’t need to be a genius to know that. Then there are some startups where you immediately think, “Okay, these guys are Olympic winners.” They’re phenomenal, they’re executing like crazy, and they have a track record. Those aren’t that hard, and we try to hire the rock stars. This is the one that I vetoed.

The hard part is what you do with the ones in the middle.

That’s always where you spend all of your energy trying to suss out: They’re not stellar, but maybe they are. Maybe they just didn’t have the go-to-market, the funding, or the support that they needed. Maybe they could succeed if we give them a chance, or maybe they’re just mediocre.

That’s where you spend a lot of your time. You have to spend time with them. You have to interview all the people. You have to have your people interview all the people. It can’t just be an Excel-sheet exercise.

Silicon Valley has a lot of lopsided companies. You’ll have a great engineering team and a bad company because of bad leadership or bad go-to-market. You can also have people who can sell anything with a ridiculously poor engineering team, and they can just sell it. You have to be very, very careful about that.

Actually, our CRO at Databricks came from a company where he could sell anything.

Yeah, he was selling SFTP—SSH File Transfer Protocol—which is free.

He was selling it for a while. He was selling it for a lot. He was making a lot of money.

We’re saying, “Electronic medical records—how important are they? If they got dropped, how much of a risk is it to your business?” Well, this is secure FTP.

You need it to be secure so somebody can’t grab that file. He’s good.

The only thing I’d add is that this strategy is probably making you more attractive to the people you want to acquire too. They don’t want to sell if they’re going to get fired right away.

Yeah, for sure. It’s very competitive.

Yeah, 100%. There’s also a reputation, right? People know. They’ll look back and say, “Okay, what happened to your previous acquisitions?”

Yeah.

Was there a huge fight and everybody’s quitting left and right, or did they work out? How are you taking care of those people? What roles do they have? Do they have influential roles in your company? That’s also important.

You’re setting a precedent in many, many ways with acquisitions and M&A: deal dynamics, the price, and what happens when you go through the lawyers and come back. When you’re spending those 20 or 30 days doing the definitive agreement, every little thing you agree to there is a precedent for the next deal.

Yep, totally.

Maybe, actually, just to turn: We’re talking about Databricks as an acquirer.

If we go back in time again to a moment when you thought about selling—and maybe you didn't actually seriously consider it—I wanted to quote this infamous email circulating around our firm that Ben sent to you.

Yeah, Ali brought it up. I had forgotten about it. You brought it up at a board meeting—you brought it to the board. But this wasn't pertaining to selling the company; it was, I think, selling a candidate, right? You talked about, “Hey, Ben, can you sell this candidate on the fact that we'll be worth $10 billion, maybe?”

The candidate was worried about the company getting acquired. He wanted a double trigger because, if Databricks sells and they fire me as a salesperson, what equity am I going to get? So give me a double trigger so I'm protected: if we get bought and I get fired, I vest all my equity immediately.

Yep. Exactly. And so, in response to this, Ben—and I'm going to paraphrase this a little bit—writes back to you, “You're severely underselling the opportunity. We are Oracle in the cloud, and we will be worth 10 times what Oracle is.” What was your reaction when you saw that? Did that give you more fortitude not to sell the company?

Yeah, Ben's crazy. I think the first thought was exactly, “Ben's crazy.” But no, I think both Ben and Mark always pushed us to think bigger.

I remember we did the pitch at a16z for, I think, our Series D, which would have been around 2017 or so. The question was asked, “What's your biggest bottleneck?” I said, “The biggest bottleneck is hiring.” He said, “Okay, well, who are you losing to?” I said, “Well, it's Google. You know, it's the FAANGs.”

The response I got back was, “Well, you need to just add Databricks to FAANG. It needs to be FAANG DB.” My reaction was to laugh. I literally said, “Yeah, yeah. I mean, this is not serious.” I was like, “Yeah, that's the problem.” He said, “No, I'm serious. You need to add Databricks to FAANG.”

Then there was a pause, and I think it was, “It's doable.” So I actually went back and thought about it a lot. I was like, “Is it doable? Am I the crazy one, or are they the crazy ones? Who's the crazy one here? Who's nuts here?” That pushed us to think about how we change our compensation philosophy.

How do we—if we wanted to go and get the best of the best out of Google—what would it require? We developed a new model. We were like, “Okay, actually, the way to think about it is your market cap divided by number of employees. That's how much money you can give away in terms of dilution.”

We calculated the number at that time, and we were like, “Wait, we're actually richer than Google in terms of how much dilution we can afford per engineer,” because, at that time—this was before the Twitter downsizing—all the companies were oversized. We did the calculations, and it turned out that we could probably pay the P95. We did the math on the P95 for engineering, and it was like, “Yeah, this actually works out.” We moved all the compensation bands and told the employees, “Okay, we're paying you P95, and we can afford it.”

That came out of that simple idea: “You're a trillion-dollar company; just add your acronym to FAANG,” and so on. Those ideas are silly and kind of crazy, but they do push you. You go back and think, “What is the fundamental reason, from first principles, that we couldn't do something like that? Why couldn't we be a trillion? What's the bottleneck to being a trillion or being part of FAANG?” Then you think about it and start zooming in on, “Can we unblock that?”

It has helped us and has been a driving force, even though it's a little annoying. It's like, “Hey, Mom and Dad, I got an A+.” “Yeah, but we ranked—” “I was number 2 in the class.” “So, was someone better than you?”

For what it's worth, when I joined the firm in 2019, the Series F of Databricks was the first deal I worked on, and I think the valuation was $6 billion. Ben said to us, “Oh, well, it's going to be a $100 billion company.” We were like, “Yeah, yeah, sure, Ben.” Lo and behold, they're doing all this work. I'm like, “What are you doing?” Like, $6 billion, $7 billion—it doesn't matter.

I was right.

He was right. Yeah, yeah. That one you have proven to be right. We still have ways to go for 2 trillion, but—

Well, the thing that you almost never get—and Ali and I had this conversation the one time we did have a real acquisition offer on the company—is this good a market opportunity with this good an entrepreneur. That's the rarest of rare things.

We see great entrepreneurs, but their market opportunity is limited. Then we see companies that have a great market opportunity, but the entrepreneur isn't big enough to fulfill that. But this was a case where we had both.

I remember, actually, the conversation that kind of flipped me. The acquisition offer was on the table. It was 6 times bigger than the valuation we had at the time, and I had made the mistake of telling my co-founders.

Yeah, they were like, “Let's go.”

They were like, “We're done.” So everyone's like, “Stop the work. Stop working. Take your hands off the keyboard. Nobody work anymore. We're done here.” Right? “Let's count my money. How much money do I have? What would you buy for that amount of money?”

They were completely not doing anything. There was just this crazy gossip going around. They had told some of the executives, and they were calling each other every day, like, “Hey, what does Ali think? You think he's in a bad mood today? You think he's going to say no?” “No, it's like, what did he say?” “He said this thing. He said this once.” So there was a lot of politicking going around, and nobody was doing any work anymore.

I was like, “Maybe they're right. Maybe we should just sell.” I remember having that conversation with Ben. I think we were in a car, both of us, and he drops the F-bombs, pisses people off, and so on, and they don't take the feedback. But actually, he did exactly the Radical Candor thing with me: He said, “Hey, you can do whatever you want. I'll support you either case.”

“Actually, if you sell for this number, it's really great for me”—me being Ben. “We make a lot of money at a16z, and I'll pay the investors back many times over. So honestly, if it's for me personally, that's probably the better option.”

“But I'm just thinking back: I was CEO of Loudcloud and Opsware, and the cards I was given—those companies weren't the company you have. When I look back, how often in life do you get a chance to even have a company like Loudcloud or Opsware, let alone a Databricks? This is just such a freaking big market. You can sell, you're going to make a lot of money, and you'll be super successful in life.”

“But if you're like me, you're going to look back the rest of your life thinking, ‘I missed that one shot. That was the one thing. I should have taken it all the way, and now I'll never know how far I could have taken it—what it could have been.’”

So, do you want to live with that, or do you want to just have the money? I'll support whatever you want to do. I really couldn't care less. I really couldn't care less.

I was like, “Okay, thanks.” I hung up. “We're never doing this. We're done. This is not happening.”

What a pep talk.

Yeah. So that's how we did it. It was excellent.

I think I also said, “I guarantee you you'll never have an idea this good again as long as you live.”

Yeah.

This is the best idea you're ever going to have.

Yeah, yeah, yeah. Well, an idea that also takes off and works, right?

Yeah.

So I want to tie one thing that you said in all of that. You were company-building, but then also just the calculus that founders, but also your employees, are making, and that's around compensation. In the early days, you could afford to pay the P95, right?

Today, there are crazy AI talent wars going on. We've talked about this a bunch throughout the summer, and we know that you can bring the best talent in the house to Databricks. How do you keep them with all of this craziness going on? Because now P95—I don't even know what that means. Is that, like, you pay a billion dollars?

Yeah. Exactly. Exactly. The joke is, which company says, “We're P50? We pay P50.” Who does that? There's no company that does that.

There is. Yeah. The 75th percentile is the single biggest lie in Silicon Valley. It's a complete fabrication, probably. Probably.

But I think it is a crazy time with AI, and I do feel bad. I actually did an exit interview with someone this morning. I feel bad for the kids right now because there's too much pressure on them. They feel like, “Oh, they have to start companies,” and I've never actually had anything like this.

Every year I talk to the interns, and I get questions about, “How do we build our own company? How do we succeed at Databricks?” The last 2 years have just been crazy. All the kids are like, “When should I become a CEO? When should I start my own company? What's a good valuation? Am I missing out if I do an internship here for 3 months at Databricks? Will I have wasted my opportunity in life?”

This is the time for AGI, and I could have been one of the guys that does superintelligence. How would you time that? How was it for you? How old were you when you were 22? What did you do?

I do think it's kind of crazy times. I also think it's exaggerated. I don't think anyone's getting $100 million offers. I mean, yes, there's Character.AI and so on, but I don't think it's actually true.

It's also in the interest of CEOs—you should know—to say, “Hey, people tried to poach Databricks people for $100 million, and they said no.” It's in our interest to say that, right? Because that kind of sets the bar at $100 million, and then any employee that cannot get half of it is going to feel really insulted.

It’s like, why don’t I get a $100 million offer? I heard on the news that other people are getting a hundred million.

By the way, Sam used that in reverse on Meta. He’s like, “Oh, yeah, they offered all our guys $100 million,” and then the next guy got the $50 million.

Now I have to pay $100 million at least. Right. That’s the smart move.

But I would say that not all startups have a valuation of $100 billion and 10,000 employees. We actually can afford to pay significant amounts, and we do pay significant amounts for the right talent.

What do you do when you’re smaller, like we were at some point? Then it’s about how big you’re going to get, what the opportunity is, what you could do together, and what it would be worth together.

Most people earlier in their careers really want to learn, and they really want to feel that they can have an impact. If you can bring them in and mentor them, stay close to them—and as a CEO, you have huge power if you can just spend 2 minutes with a kid out of school. It’s immense to them if you say, “Hey, I’ll even mentor you. I’ll help you. What do you want to do in 5 years?”

“I’m thinking about starting my own company, actually, in 6 months. I’ll work at Databricks for 10 years, but in 6 months, I would love to be a CEO.” Then you can say, “I can coach you. I know how fundraising works, I know the early days, and so on.” You can mentor a lot of them, and that’s actually worth a lot to them as well.

But in general, help them be successful and help them build their careers. If you’ve done it before, like we have, you can calm them down a little bit and say, “Hey, you have a few decades. Don’t worry about it.” The FOMO and the pressure have to be reduced, and I think that’s also calming. They feel good about it.

Yeah. Yeah. I always say the best cure for starting-your-own-company fever is to start your own company, and that’ll teach you.

It’s not that easy.

By the way, they often come back to Databricks after starting companies, and they’re much more thankful. You understand—and actually, I didn’t mention this earlier when you asked about acquisitions—my favorite acquisition starts with the people, right, and then the product.

With the people, I love to hire people who have seen what it’s like at a big company. I don’t know if it’s great, but they’ve seen processes scale at a big company. They’ve been at Google or Amazon; they understand the processes, so they understand how to navigate a bureaucracy and work with it, and they’re not just going to be inundated by it.

But then they’ve gone on and done their own startup, and that’s really, really hard. It’s extremely hard trying to do everything yourself, and you don’t have any help. You’re trying to do this in a crazy market, and you’re trying to compete with $100 million offers when you have nothing.

That takes a certain amount of grit, and it’s really humbling. I love the people who’ve done both of those. They end up being the perfect employees at Databricks because they come in and they’re really thankful. They’re like, “Hey, what these guys have done at Databricks is actually really, really hard. I tried it, and I’m really good. I was one of the best at Google or somewhere, and then I did my own startup, and we absolutely failed. So, hey, show some respect here. These guys know what they’re talking about.”

So those are great employees, actually. I think you should keep a great relationship with people who leave your company, because they can boomerang back in a couple of years.

Yeah. Yeah. And look, it’s very hard to make these things work.

It also requires a lot of luck. I think one of the things people don’t realize is that a lot of things have to go right that should never go right.

And a lot of things will go wrong, but if you can grab your lucky moments, that’s a rare thing.

One way to prove that is this: Databricks started in 2013. If we had started in 2012, that rocky year—that difficult year—2015 would have happened in 2014, right? To start, we didn’t have the revenue, but we were a cloud AI open-source company. Those things didn’t take off in 2014, so even if we’d had to do the CEO change and all of that, and I had become CEO a year earlier, we were too early in the market. The cloud hadn’t taken off. AI wasn’t even a phrase; AI meant robotics. People used “machine learning” as the phrase, and the company would have failed. We wouldn’t have had enough momentum. There wasn’t enough cloud TAM to be had.

If we’d started the company in 2014, a year later than we actually did, we would have had our difficult year in 2016. But by 2016, the cloud was starting to happen, AI was starting to happen. We would have done the fixes in 2017, and it would have been too late to the party. The hyperscalers and our competitors would have taken it away, and we just wouldn’t have gotten enough momentum to succeed. That’s the timing of when we started.

So how did we clock it so well? We had to wait for Matei to finish his PhD thesis. That’s it.

Yeah, that was the whole thing.

So there’s a lot of randomness, and you have to get lucky.

And it was so on the edge as it was. On the Series C, Ion had a handshake with Redpoint, and Redpoint just stopped returning his calls, to the point where the Series C was led by a16z, which also led the Series A, and NEA, which led the Series B. We co-led the Series C because nobody else would do it. I mean, it was that close to going under. Most companies wouldn’t have made it; that would have been it.

Yeah, it was very close because we couldn’t get funding from anyone. Funding froze up, and nobody wanted to invest anymore. So it was really a lifeline from a16z.

Yeah. We were just burning a lot of cash. We weren’t generating much revenue other than Spark Summit.

We had a lot of downloads.

A lot of downloads.

And recurring conference revenue.

Yeah, and recurring.

How confident were you at that time, when things were at their lowest?

I seriously considered taking the professor job at Berkeley because I seriously thought this was going to be very, very hard to pull off.

The sentiment at Databricks—or at least my sentiment—was, “Look, you win some things, you lose some things in life. We created Apache Spark and made it a worldwide sensation. Everybody’s downloading it; the downloads are through the roof. We have this great conference—thousands of people come to our conference. It’s awesome. Let’s go back. Let’s do it again. Let’s publish another paper and do those kinds of things.”

We’re just not business guys. We don’t understand business. That’s okay. We don’t want to be business guys. That’s kind of how I felt about it, right? But what I knew was that—

By the way, everyone went back and became a professor. All this stuff happened.

Yeah. Ion went back, and in 2015, we knew that we had tried everything.

By the way, PLG—product-led growth—was something that we had tried very hard, and it didn’t work for us. Actually, one of our biggest failures was PLG at Databricks. Everybody kept telling us, “PLG, PLG, PLG.” We were like, “Okay, product-led growth. Amazon isn’t just going to swipe their credit cards. We don’t need salespeople.”

Except Cranny.

Exactly. Yes, that is true. Kudos to Mark.

So that year, we had formed some hypotheses. We had nothing to lose. What if we just pivoted these things? What if we went all in on B2B enterprise sales? Certainly, PLG wasn’t working. At $3 million ARR, that’s not going to take you anywhere.

They were just taking our open-source software, so we had to have proprietary code around it. The executive team were all PhDs, so what if we brought in someone who didn’t have a PhD and saw how it went?

I’ll never forget Arsalan going, “We made the number.” I was like, “You made a ridiculous number.” You made the number. If you keep making that number, you’re going to go bankrupt. You didn’t make the number. You made a number that you set that was way too low. We haven’t figured it out.

Ben was very nice and complimentary in our board meetings that year, 2015. We were in a bit of trouble. Let’s say it was very truthful.

But, yeah, we had nothing to lose. We didn’t know that we were going to succeed, but we had nothing to lose by making those big changes, and we made them in 2016.

It turned out those were the bottlenecks: giving away your software for free, not having executives who had seen the movie before, like Ron, who came in, and the PLG motion not being enough. So maybe we should just try. We weren’t certain that B2B would work, but we knew that PLG wasn’t working for sure.

Yeah. Well, another thing is that we got Ron. The fact that the first sales guy we hired was a sales savant, a genius—that never happens. He was a guy we didn't know. Our talent team found him from some company we'd never heard of.

Yeah, a French company.

Really, the only reason we hired him was because he was the only guy Cranny ever liked. Of all the sales guys he ever interviewed, Andy was like, “This is the guy.”

Wow. He’s a new-generation Marc, but we just stumbled into him. Unbelievable.

Without Ron—

It’s very hard to see this company getting to where it got to.

There’s some luck involved in even finding him. But he was phenomenal, and kudos to Ion, who actually led the search in 2015. Ron was game-changing for us, but he was a very uncomfortable hire.

Because he did not have a PhD—

He did have an engineering degree. He has an engineering degree from Stanford, which helped a little bit, but he’s a sales guy through and through. He’s a classic salesperson who grew up in sales, even though he has an engineering degree.

The comfortable thing would have been to pick someone—and we had some candidates in the mix—who were supertechnical and using the product, giving us feedback. That would have been much more comfortable for us.

Yeah, Ron was uncomfortable. He was a very uncomfortable hire, and he made it very uncomfortable for us for many years. He still does.

But that’s a lot of the key to the company.

It forces a customer focus that would be impossible to have without somebody that smart and crafty about getting his way. I mean, just unbelievable.

If you can also keep the original team together, that’s important. We were 7 co-founders still. Many of the co-founders—you said data warehousing was a big push for us. My co-founder Reynold was really the one who pushed this.

The contribution level from a large number of co-founders is unique in the industry. You’ve got Patrick, you have Reynold, you have Matei, you have Arsalan. It’s crazy how much—

The original team contributes.

Yeah.

The PhDs all contribute. Arsalan really made the go-to-market work, and he really made Ron work with the rest of the company. That was supercritical. Matei continued doing lots of innovations over the years. Patrick led all of engineering in big chunks of it, and so on. We’ve been lucky to get such folks. Hiring is critical, and keeping the original talent—I think those were some of the things.

Usually, only 1 of the co-founders contributes long-term. To have that going, and to have Ion still on the board and Scott still on the board, I mean, it’s very unusual.

Yeah, we have a lot more we can get into, but we’re at time, so we’ll leave it for future episodes of Boss Talk, but this is a great first episode.