The Exchanges

Q Walk me through the product or really the unit of what that model driven approach means. Like conceptually, it's not that hard to understand, but what were you actually providing to all these companies?

A What Domino provides is basically the system of record for data science. So if you think about how HR uses Workday and salespeople use Salesforce, the thought is that data scientists should use Domino. And so what it does depends on where you sit, much like those other applications, what they do depend on where you sit. If you're a data scientist, it's a workbench that gives you access to compute. It gives you access to revisioning of your data and your models and your results. And those are probably the two kind of big things it does for you. And so basically just creates a good environment for which you can do data science. But if you're an analytics manager, essentially it gives you a repository of all the models in your firm. And it gives you a sense of what everybody is working on in the firm. And so then it lets you drive productivity by getting people on board faster and finding things faster and sharing knowledge faster. And it also lets you monitor those models. So are they decaying? Are they, are they working the way you expect? Are they going off the road? But all that comes from the idea that every model is getting built in Domino and stored in Domino. And so then basically if you're a data scientist, you get those tools. If you're an analytics manager, You get the ability to monitor it. And then as a company, you're able to access it all. So you can hit the models …

AI assessment note: “What Domino provides is basically the system of record for data science.”

Q What's an example, just one granular example, of that communication link working where it may not in a, say, a different organization?

A I'll give a couple of examples. So last year, We were looking at Etsy and some of the top line numbers in Etsy didn't look great. And so, like I said, we have all these charts and tables and you sort of look at that we're predicting user growth and these sorts of things. And things seemed a little bit below consensus, but one of the analysts said, Hey, like, can we do more here? And the data scientists team was able to essentially cohort all the data based on how long a person had been using Etsy. And to basically kind of create like the most loyal cohorts versus the least loyal cohorts. And when you looked at it that way, what you saw is that the loyal cohorts, their revenue growth was actually really expanding, was expanding pretty rapidly. And so you sort of had this core group of users that mattered and they were really sticking with the platform and doing really well. And then you had what I would almost call like noise in the data that Etsy had made some acquisitions and the acquisitions were sort of People have gotten grouped in as users and all this sort of stuff. And so that's the kind of thing where an investor appointing to has a great understanding of the company is looking at that data and then has a question and the data scientist is able to then use their technical capabilities and all these data sets that we have. And so, by the way, to answer something like tha…

AI assessment note: “So last year, We were looking at Etsy and some of the top line numbers”

Q If you drill down on that, were there one or two implementable improvements that you saw in your time there that made a meaningful change in the organization?

A Well, I'd say a lot of it was around the structuring of roles and the clarification of who did what and what their responsibilities were. You have a project and you're trying to figure out whatever it is, how to estimate growth better or estimate inflation better. Making sure that the team around that has all the skill sets that are needed so that they're working as efficiently as possible and so that the people's time at the top is being used in as smart a way as possible. So, you know, you can imagine that you want to have the data people and you want to have clean data that's accessible and you want to have that ready to go before you start doing your basic modeling. Before that, you want to make sure that the questions that are being answered and asked are structured. And so that was probably the most important thing. And some level it sounds very basic, but a lot of research teams don't run with a lot of structure. They run more like a five-year-old soccer game to kind of way, you know, everybody chases the ball and we were trying to scale something. And so that was one of the key thoughts we had.

AI assessment note: “a lot of it was around the structuring of roles and the clarification”

Q When you're behind the scenes and you're able to look at all the data of the success of individuals, to what extent do you find that, you know, start on that first bucket of the person plus machine that someone who comes in, maybe an experienced investor from the outside, how much are they able to get better from using all of these data driven inputs?

A I think in some sectors, it's almost impossible to trade the sector without a lot of the data inputs. It ties a little bit to one of the trends you see in the industry overall, which is, I think the strength of the platforms and the power of the platforms. And basically we have the economies of scale to be able to afford millions and millions and millions of dollars of data. And so if you look in consumer as an example, like I don't know how, if you're only managing like seven or eight hundred million bucks, like how you even compete against Our team plus the teams at the other big platforms. In other sectors, I would say some of the alternative data in particular is newer. So there, there is more differentiation. We don't run pure A-B tests in quite the way that your question was suggesting, but I would say that the data is helpful, but it is just one input, and the real secret sauce is the ability to Meld as many inputs as possible into a recommendation. And I think that's what makes our investors really, really good. And then what we try to do is give them as many inputs as possible. And this is just another set of inputs that in, like I said, in some sectors are really, really hard to compete without.

AI assessment note: “in some sectors, it's almost impossible to trade the sector without a lot of the data inputs”

Q What are those theses in the bank side?

A The main thesis on the bank side is that You've had all these quote unquote fintech disruptors come along, and there's been some good businesses built out of that, but mostly what those companies are is they have a tech advantage, but they don't have access to customer advantage, and they don't have any sort of cost of capital advantage like a bank does and things like that, and so they've built technologies that people like, but if we were to sort of offer a very broad brush observation about all of fintech, they've often had a hard time making their economics work. And so our thesis is more that Now banks and other financial providers have to respond to the improved tech through digital transformation. So we've made a lot of investments in all the different things that a big bank like JP Morgan or whoever is going to need to, well, not just big banks, small, medium, big. Good thing in the U.S. is there's lots and lots of banks, lots and lots of financial service providers that they're going to need to be able to drive that digital transformation in the way that customers want. I mean, consistently in surveys, people always say that their financial service providers have the weakest technology, and we just don't think that that's going to endure. So that's probably our biggest thesis. Globally, we also have a thesis that is that around this idea that the distinction between fi…

AI assessment note: “The main thesis on the bank side is that You've had all these quote”

Q And what did you find in that process of trying to learn how to scale the research effort?

A It's kind of like almost any process that you're designing. You have bottlenecks. One of the great bottlenecks in research is the time of The most important and most creative researchers. And so the metaphor we often use was an operating room that we were sort of trying to set it up so that a Greg or a Ray or, or Bob Prince could come in and be maximally impactful. And so part of that is around the technology. Part of that's around the data. And then part of that's around having two or three different types of researchers who did different types of work. But ultimately, you want those experiences to be really impactful and get the most out of their time. And so that's probably one of the central metaphors we used when we were thinking about it.

AI assessment note: “One of the great bottlenecks in research is the time of The most important”

Q It's hard to talk to someone who's had any duration at Bridgewater and not try to ask some question about the culture and your impressions of it. So what did you take away from that experience?

A I joined when I was still fairly young. I think I joined Bridgewater when I was 27 or so, and It never struck me as that weird or strange or all the headlines that kind of get produced about it or whatever. I mean, the, the fundamental tenants of set strong, aggressive goals, be transparent, be honest with each other. Like in my mind, they're not particularly controversial things. And I never found it a particularly controversial way to be. I mean, in fact, in a lot of ways, I found it a very simple way to be. If you had something on your mind, you said it, and if you, if you wanted to say something about somebody, you said it to them, not to someone else, and so I always enjoyed the culture, and after I left Bridgewater and started my own company with two other Bridgewater people, we took a lot of that essence with us. I'd say the difference was that Bridgewater has a lot of formality around it, and there's a lot of books, and there's apps, and there's all these things that exist because they've been so successful, but that core just kind of like way of being with each other When I founded Domino Data Labs, we took all that with us and do all that to this day. And again, though, I think it's largely because it's kind of like how we were as people, and then we saw how effective it made the environment that we were in.

AI assessment note: “When I founded Domino Data Labs, we took all that with us”

Q There's a lot of discussion of the use of data and computers and computer driven trading, particularly in shorter term investing. And I'm just curious from an outside perspective of .70 to being an organization that turns over portfolios and a lot of those multi-manager structures tend to be that way. What's your perspective on this kind of notion of how computers are fitting into the investment landscape?

A .2 is a really interesting place to work with regard to that question, because the firm is very well known for its history and discretionary investing, but Steve's also had systematic teams since. So almost as long as anyone. And, you know, I would say one of our major theses of the firm right now in terms of our strategy going forward is that The world of discretionary investing and the world of systematic investing has been, has been kept too far apart. At some point, discretionary investing has obviously gone on for a very long time. Quantum investing rose up, but it sort of took its own path, but that each of them has sort of different things to offer the other. And so a lot of what we're trying to figure out is how to bring those two worlds together in terms of what they each offer. And, you know, I think where that starts, you know, is really kind of this question of what people are good at versus what machines are good at. When our thesis is that humans are particularly good at the creativity required for idea generation, good at working off of thin data or very scarce data, good at defining and refining goals, knowing what questions to ask, seeing the big picture and how things have changed and how regimes are shifting, and good at emotional intelligence. To reading a room, talking to a CEO, those kinds of things. Whereas computers are very good at math. Scale. Repetiti…

AI assessment note: “our thesis is that humans are particularly good at the creativity... Whereas computers are very good at math”

Q Matthew, this is sort of baffling to me that you went from whatever you were doing earlier in your career, and all of a sudden a macroeconomist research at Bridgewater, and now all of a sudden you're a venture investor. What did you learn in your first couple of years of running .72 Ventures that you didn't or couldn't have known before you were actually in it?

A We went in with this thesis that we would be able to figure out on some multi-year timeline How the world was going to change and, and generate investment themes in that way. And that was kind of a thought. The question is, could that really work? Could you really get experts that were able to see that and do that in that way? And, you know, I think that our answer is like, yes, actually you can do that. And if you take like a thesis, like our modern thesis, or in FinTech, we have a bunch of theses around how banks are changing and what they need as a result of that. And you put all that together, and I think that the number one thing I've learned is that that is actually a doable thing with the right experts.

AI assessment note: “the number one thing I've learned is that that is actually a doable thing”

Q Where do you think this business, and when I say this, it could be hedge fund, venture, private equity, hyperscale, all of it. Where do you think this business looks like in five years?

A I think the big will get bigger. I think that there are definite advantages to economies of scale. Like, you know, I think about how much we invest in our risk systems, our trading systems, our data systems, our research systems, our training, all these sorts of things. And that All of those things create advantage and they come from being at a certain size and having certain economies of scale. And I think being a small fund is harder and harder. I think that as we talked about earlier, I think you'll see quant and discretionary become closer and closer together. And I'm not sure if it's like a full merger or not, but they'll certainly be more similar than they are today. I think you'll see more firms investing across the capital structure. So doing public markets and pre IPO and venture in part, if you've done work and you have insights around automobiles, let's go back three or four years. You know, you want to, you want to invest in cruise. You want to invest in Tesla. You want to think about GM. You want to think about Uber. You want to think about Google. You want to think about Lyft. And so most of those companies now are public, but four or five years ago, they weren't. So you want to be able to sort of use those insights across the whole spectrum.

AI assessment note: “I think the big will get bigger. I think that there are definite advantages”

Q So let's dive in. We can start on the hedge fund side. .72 and its predecessor, this isn't the new organization in many ways. So as you were coming in, what did you initially do?

A It's not a new organization, but you know, one of the things that I think makes Steve just really remarkable at what he does in this industry and probably in the industry to be in is just The restlessness and the dissatisfaction with everything that is and this emphasis on what needs to be. I still remember the very first time I met him, my first question to him was, I said, you have built a more enduring hedge fund than almost anyone else, right? I mean, there's five or six people like that. Why do you think that is? And he said, because I burned it to the ground three times before and I'm getting ready to do it again. And you know, what he meant was that what is today is not what's going to be profitable tomorrow. And you have to constantly be changing the business. And he has that restlessness deep inside of him. And so the very first thing I did was I took over responsibility for what's called our proprietary research. So Steve had a thesis that we really needed to build a lot of in-house capabilities around research. And that we needed to be able to essentially generate a lot of our own information about the companies we were investing in. And some of that had to do with how do you best produce those insights. Some of that had to do with the growing explosion of alternative data. Some of it had to do with changes in the sell side and how potent that information was and how…

AI assessment note: “the very first thing I did was I took over responsibility for what's called our proprietary research”

Q And so how does that get disseminated across the many portfolio management teams?

A There's two different answers to that, and so one answer to that is that we have different product lines, and the product lines essentially vary based on How much technical acumen a team has, and so we have a lot of teams that have analysts that are fairly technical, and those teams plug directly into our databases and things like that, and can pull the data into Excel models and manipulate it and all those sorts of things. On the other extreme, we produce a lot of reports that would read to you like a A fairly dense cell side report. And so, you know, they're not written quite the same way, but they're written reports that you can read. And then in between, we have all sorts of tables and graphs and those sorts of things that we produce for the teams. So that's kind of one, that's the product vector. The other approach that we take, though, is we do a lot of bespoke work, and we work really closely with the teams. A lot of times I'll be on panels and things like that, and this whole theme about where does the competitive advantage in data come from, and all these sorts of things comes up. And I actually tend to think it has very little to do with the data sources itself or the information. It's the competitive advantage you're looking for is the ability to integrate across your investors with your data scientist, with the people who are sourcing the data and get that whole cha…

AI assessment note: “we have different product lines, and the product lines essentially vary based on How much technical acumen”