Invest with AI
00:01:17 · Kirk McKeown
[00:01:17] S0: Yeah, sure. So I started working on Wall Street in 99, uh, just to just to put some put a pin in the date, um, where I went to tutor, I was at Tudor Investments, um, and I was working in the private equity group, the venture capital group, uh, sourcing deals. So I cold called for the first number of years of my career. Um, you know, sort of and did research on deals. Um, it was based in Boston. I rolled up to uh Carmen Scarpa, Bob Forenza, and the venture group, uh, and Jimmy Pilata ran that office. It was it was part of the public equities framework, a tutor. And then I ended up at Glenview, where I worked uh for Larry Robbins for six and a half years uh and built out the primary research business there. Uh what that meant was um I got tapped to build a call center business for all intents and purposes, uh doing in you know, sort of systematic industrialized channel checks uh across uh hundreds of supply chains back in 2007. Um that meant uh we were doing you know, sort of thousands and thousands of calls uh with uh C-suite executives to better understand what was going on on the ground and everything from HVAC to flooring to manage care to consumer retail, uh, tracking change systematically on a monthly basis. And I was doing a couple thousand calls a year myself, right? So I had this really cool seat and vantage point um to learn about all of these market structures and business models and companies and brands and businesses. And in 2012, I ended up uh heading over to SAC Capital uh to build Canvas and uh Canvas's uh the calls business at 0.72. And then over the next nine years, I was able to be involved in all facets of the proprietary research build and run. Um and sort of uh when I left uh in 2021 to start CarbonArc, I was managing you know some market intelligence uh and Canvas. Uh so all of the calls and all of the data that was servicing the discretionary long short businesses at 0.72.
00:03:14 · Brett Caughran
[00:03:14] S1: What um can you can you explain a little bit more what you said systematic and industrialized channel checks? And you know, I could I'm looking at a company, I could call one or two people in the supply chain and sort of how do you how do you then distill those insights back into an input into the financial model? Can you educate us a little bit on just how do you system systematize and industrialize it? Like how do you get a big enough N of calls? How do you correlate those insights into an actual investment insight? Like how you how how did you think about the so what? Like here's a bunch of raw call transcripts. How do you translate that back into an investable insight in a financial model or an investment thesis?
00:03:54 · Kirk McKeown
[00:03:54] S0: Absolutely. So, you know, sort of when I when I when I actually was offered the opportunity to do to do that in 2007, I started as an analyst working for Larry. I was covering consumer and tech in the US and Europe. Um, you know, sort of I actually was in Europe when the group started. And at the time, you know, sort of I got tapped because I was really good at the primary research stuff. I had taken, you know, sort of basically as a backstory, I had I had taken all of my experience in generating cold calls when I was at um at Tudor, and I had deployed it into my interview process to get into a hedge fund when I was coming out of business school. And that's what ended up getting me my job eventually at Glenview, was I had done a big project on Sears and Kmart in 2005 when they merged, and I did hundreds of calls around the name and I interviewed around the street on it, right? And so I had sort of made my name early days during interviewing process on that project. Um, but when when I got tapped by Larry to build that group, I didn't actually want to do it. It's the funniest thing. It was the the the single most important call of my career. And I wanted because I wanted to be a quarterback, you know what I mean? I wanted to be an analyst, and then I wanted to be a PM, and then I wanted to, you know, sort of be the guy. And um, and then, you know, sort of uh I ended up reading the blind side, um, which talked about the left tackle position, and nobody knows their name, but the highest played player on the field. And I was like, okay, if I'm gonna be a left tackle, I'm gonna be the best left tackle in the business. And so I I I you know started building the primary research business at Glenview. And I the first thing I did was like really understand how I was going to compete with everybody else. And so I looked at the field. When I looked at the field, um, you look at an analyst at a in a shop across the street, they were likely doing two to five calls at the end of a quarter to check a thesis or uh, you know, check on sort of math that they believed to be the case for their companies. Um, they were likely the guys doing the work on the name. And so, you know, sort of by setting up a group inside of Glenview where we didn't, I didn't have PL, I wasn't paid on PL. I was focused on getting to ground truth in the best way possible.
00:06:09 · Kirk McKeown
We set up an ecosystem that was actually better for confirmation bias and compliance right off the rip. Um, and then it was really about, okay, if they're doing four calls, I'm gonna do 20. Right. And if they're gonna do them quarterly, I'm gonna do them monthly. So their four calls are my 60 calls, and I'm already talking to their four people, so I have a better sample. And then it was really around once you had sort of the the construct around the size of sample, which isn't statistically significant, by the way. Um, you know, these are these are these are sort of flow-based checks. What fundamentally happened was it was really about developing a system to track change in those, uh, to track change in those supply chains to not only have a better understanding of the longitudinal veracity of those checks, um, but also to understand sort of rate of change first derivative and second derivative. And so, you know, what came out of it over years and years of doing it was really devising a system that allowed for um, you know, sort of uh systematic collection on 45-minute calls with pre-approved compliance scripts, um, constantly improving the size and scope of the companies throughout the value chains, and then sort of tracking change through time and then just being really deliberate and intentional around uh racking your wins and losses. Um, you know, sort of by collecting a lot of information on these calls in a legal and compliant way, you can really start to understand based on first derivative and second derivative changes of the key KPIs in the ecosystems, what's actually going on? Lead times, build schedule, shortages, surpluses, um, anecdotal collection, and then asking the last time they saw it to create some reference point from the past. And then what happened in that framework? So we ended up building and developing this really uh this really cohesive interview process, collection process, and then synthesis process that frankly, um, you know, sort of I still think is is is the bellwether for you know, sort of what I would call sort of supply chain checks. Um never never tracking uh a specific name, but actually looking at cabinets or flooring or mattresses, and then understanding through that how you sort of market structure, business model, and near-term catalysts are impacting flow through the channel, and then looking at business momentum.
00:08:25 · Kirk McKeown, Brett Caughran
And, you know, sort of I I you know I could talk about it all day, but like it it really came down to being general about the specific, specific about the general, and avoiding false precision. Right. Um, and and those are like key tenets in that type of work. Uh everybody wants nine decimal places to feel safe. Um, but and at the end of the day, everything in life is better, same, worse than expectations. And if you start there and keep it simple, you can really start to create some space around understanding over time how things are forming and shaping in different ecosystems. [00:08:59] S1: What uh can we talk a little bit about the craft of just conducting the interviews? You've seen this sort of big rise of expert network calls, and now the hot thing is AI-driven expert network calls. Um, and sort of the assumption of many is like this is a monolithic process. You just get on the phone with an expert and all the sort of information will sort of flood into you. What's what's uh talk talk to talk to me a little bit about just the artistry and the craft of framing the question correctly and like what does a good interviewer do in these calls that a poor one wouldn't?
00:09:33 · Kirk McKeown
[00:09:33] S0: You know, it's it's so two things. I always I always let people listen to my calls when I was doing calls. And then the first thing I'd say, listen to my call, and then never run a call like I do, right? Um, because I would I was actually a high frequency trader of calls. I would actually jump on 25 calls in a day, do 10 minute calls, collect anecdotes from 25 people, and then sort of create a portfolio of inputs around that based on sort of the ecosystem. I wouldn't suggest that to a team that was doing these calls. Um, those are usually 45-minute calls that is sort of both, you know, what I would call knowledge and trend, right? At the end of the day, there are two types of checks in my eyes. There's a knowledge check, which is how does Databricks make money, right? And then there's a trend check, which is how are Snowflake and Databricks competing against each other in the SMB market? Right? Knowledge versus trend, right? Um, and the best checks have both. You end up working both of those into the call. But fundamentally, it's really about taking the taking the the consultant on the journey and really building an understanding of where what what their vantage point is in the ecosystem, because that's the most important thing. What can they see from their seat? And then that drives what the most important part of the conversation could be. If you're talking to a paint contractor about the paint supply chain, you don't care about their demand. Like what's happening if they're a paint contractor in San Francisco, all you care about are lead times from Sherwin-Williams and price points, you know, sort of from the distributor. So you got to figure out the scale point in every conversation and then really form and craft and understand the believability or credibility of the person on the phone, right? So there's an optimization around, because that goes into weighting, because that's really important here. I think one of the challenges that you're gonna have with the agentic workflows over time until they get sort of um, you know, sort of richer and more textured, uh, is weighting data points against the person who's telling you the data points, right?