Michael Dempsey - Compound VC
We talk about public market vs private market investing, machine learning, SPACs, Total Addressable Markets (TAMs) and more!
Michael Dempsey is a General Partner at Compound, an NYC based VC firm that invests in bleeding-edge technology. Previously Michael worked at CB Insights and Crane Partners, a multi-strategy hedge fund.
We talk about:
— Public market vs private market investing
— Machine learning
— SPACs
— Total Addressable Markets (TAMs) and more!
NA: Could you give us a very quick overview of your background?
MD: I’m born in Los Angeles, grew up in Boston, Massachusetts and eventually came to New York to go to school. I really hated school generally. I got out as quickly as possible and went to work at a hedge fund. The hedge fund did a bunch of different types of investing from public markets to long short and volatility-related trading, to private markets. I eventually led a small carve out to do seed-stage investing along with the other things we’re doing.
In that time I really fell in love with it compared to every other type of investing I’d ever done. I then realized that if I wanted to do that, I should probably figure out how startups actually work. I joined a company called CB Insights pretty early on that analyzes data exhaust around private markets, startups, venture investors, etc, and helped build out a lot of the data and research team there. The company is most well known for the newsletter, but it’s actually an enterprise SaaS product. At that time I saw the company really grow from a small to large team.
In that time, I really started to figure out two things about the venture world. The first was, the lack of deep independent conviction in research that has been done and contrasting that to the hedge fund world, it’s pretty materially different. Hedge funds its all about information asymmetry, the venture world is very much not that which felt kind of broken to me.
Then secondarily, I really enjoy the deep tech and frontier tech type areas. I spent a lot of time building a brand there, and eventually made my way over to the venture side and now have been at Compound since 2016.
NA: How different is investing in the public markets compared to the private markets?
MD: Yeah, it’s wildly different right. I think with public markets, the thing that I didn’t love is, you feel like you could get picked off at any time by someone who’s just working more than you and I think that’s still true in venture, but it feels likes you never feel totally, okay learning You always feel like you have to rush through everything.
I really didn’t like that. I will say the discipline and some of the rigour that’s applied to public market investing has helped me on the private side where I think, especially as a seed-stage investor it was a pretty different type of rigour, I guess I would say.
Having that understanding of how you quantitatively analyze a business, I think does help even just for helping portfolio companies as they scale past the two cofounders and an idea that we invest in sometimes.
NA: You mentioned, machine learning a bit. How would you talk about machine learning (ML) to an investor who is just starting to look into it now? Where do you see the true applications?
MD: I split them into applied ML companies and infrastructure companies and infrastructure is super hot right now, everyone wants to do it. I think a lot of it spawns from this idea of how hot developer tools as a market is, the difficult part within that and I’ve talked to a lot of investors about this who have asked about infrastructure, within developer tools, developers know generally what they want and what their workflow is, and there’s a system where tons of people are pushing code to production systems.
In machine learning, not that many people use true machine learning models in production software and there isn’t any agreement upon a machine learning engineer, data scientists, data analysts, or researcher. These titles are blurred, depending on where you are and what company you’re at somebody who’s a data scientist, one place might be considered a machine learning researcher at the other place. There’s no consensus on what part of the stack is broken and what they need to fix.
Often when I talk to new investors, I say on the infrastructure side, it’s really hard to invest in, we’ve only made one investment in the space, there are some firms that are incredible at it and more props to them.
On the applied side I think the big thing that people get hung up on is the canonical question, why won’t Google, Facebook, Amazon, etc do this? The first answer is they just don’t ship product. I never worry about Google fundamentally destroying most businesses, because they don’t ship, ever. When they do, they then kill it six to eight months later. Awesome company for big reasons. but I don’t worry about that.
The other part, which is the commoditization curve of machine learning is probably materially faster than anyone thinks and definitely more than anyone thought. So if you looked at how people were thinking about deep learning companies years ago, there was a strong view, that if a company has the best algorithm, they’re incredible. Then it became clear that the state of the art machinery gets destroyed every three to six months. So you can’t really bet on the best algorithm company.
You have to think of how does machine learning work within a given product, maybe how can you have a data flywheel that makes your models more defensible? Maybe there’s a certain type of category that allows you to function better and then build software around the expertise of machine learning that serves more of a full value stack. There’s a lot of nuances, on the category and the customer, but I think the biggest thing people that haven’t spent time in the space get hung up on is:
“Oh that amazing team wrote this paper that everyone loves, we should back them to build a startup, to do machine learning stuff in the startup.” The reality is that they better have way more thinking around what the company is because either of those big companies is going to destroy them because they can throw more compute at any given problem.
Or some other random group of researchers is going to disrupt them in three months from now with a new paper and no one can reproduce and so it’s a disastrous scenario. I think more and more applied machine learning companies are as much product-centric companies as they are technology companies and that really matters.
NA: Interesting perspective. I recently read your write up on SPACs and thought that was really interesting. Could you talk about your general thoughts on SPACs and how do you actually feel about them, being a seed investor and seeing them as a new exit mechanism for a company in your portfolio?
MD: Yeah, as a seed investor, I think they’re great because anything that allows companies to get liquidity is great. I also think that the way in which many of these SPACs have been priced, frankly, is kinda crazy. Again, as an early investor, who isn’t buying into those, but maybe would be selling to them, more power to them.
I think that the broader trend, and I actually don’t know if my public market experience actually holds weight in this situation, but I think the broader trend is people across all asset classes continually understand that private market are eating returns and there’s a secondary argument that you can make, which is, that’s not true if you look at Shopify and some of the companies have grown massively after going public.
Sure you can say that, but look at the series of seed investors in Shopify also. So there’s a, there’s a counterpoint to all these statements. I think as a concept SPACs are good for some types of companies. and also for the continual trend that public market investors want to get access to private market type risk and private market type utilities
I’m worried that VCs and portfolio of founders, not all of them, but some of them of like middle high-end companies that probably couldn’t IPO successfully think to SPAC them. You then are getting some form of adverse selection or some form of companies that on the plus side that the public market normally wouldn’t be able to properly understand, but because of the brand signalling of other investors associated with the SPAC, investors are willing to stomach like weird numbers over your dynamics within the company for a longer period of time.
NA: Recently I’ve read a few articles and thoughts from VCs talking about the total addressable market. One of them was referring to Bessemers Shopify Memo and how they thought Shopify’s TAM would be $400 million and now it’s well past that. After seeing this, how realistic is it to look at TAMs and how does that affect decision making?
MD: I don’t think about doing TAM analysis almost ever at the seed stage. It’s binary for me, is this a massive market or not?
I will say, our positioning as a smaller fund allows us to not have to need $50 billion companies to return our fund. We are pretty happy with smaller level outcomes, still billion dollars plus type things, so the burden is a little different.
One of the smartest things, everyone at the time didn’t think so was, Andreessen taking a very strong view in 2011 when he said software is eating the world. There’s an implicit view in the way in which they were operating, is that the addressable market and the scale of these companies is way underestimated.
So while everyone is trying to price them in certain ways, you might think we’re overpricing it, but we think we’re still underpricing it by blowing everyone out of the water by 20 to 80% on however we price these deals. I think that view is now way more commonly held within venture and tech ecosystems now.
I would say I’m just not the type of investor that’s gonna do a deep analysis on what is the actual addressable market it’s kind of like, is this big enough to be a massive, a hundred million dollars plus a year business? or not? And if it is, cool we’ll see what else happens five to seven years from now.
NA: I feel entrepreneurs who are making a pitch deck, feel like they’re compelled to have a slide talking about TAMs.
MD: I think it depends on the category. At the seed stage, I think it’s less important. If you’re doing growth stage stuff, that matters a lot more, right. You need to understand how much room has this thing left to run. What’s the market penetration of that given piece of technology?
NA: My final question is what’s the latest publicly announced investment you’ve made and why did you make it?
MD: The most recent one that we did publicly announced, which was made a while ago, is a company that’s actually started by my partner at the time. He’s now a venture partner with us, Josh Nussbaum called Halcyon Health.
We invested in Halcyon because as a firm, we spent years looking at broader substance abuse, issues within the United States and globally and the health care system.
We had previously made another investment in a company that ended up going out of business that was using AI to try and treat addiction using intervention therapy. One of the things we saw was that and I think we’ve seen this across a couple of our health care investments, there are certain levels of technology that will create theoretically better experiences and better outcomes, but often you really need to own more of the stack than you think.
With the AI for addiction company, we saw that it definitely impacted some level of outcomes. The usage of it, the willingness to pay was less clear because you couldn’t feel a lot of the outcomes in the short term and the changes in the short term to your behaviour as someone who’s addicted to try different things as a user if that was your user base.
Josh, as he spent more and more time with the company, as it ended up going out of business, really started to understand that the thing that was broken with a lot of the substance abuse issues and care models is that you would get into these one size fits all models and you would get brought into the system that you can come in through the yard, you go to some form of treatment and then you’re just kinda like lost. Your spit out and you are lost.
For us, it was really about seeing that not work once and seeing all the problems that exist today, but honestly, the economic value and economic destruction that those problems create, along with the human side. Josh has been very open about this. He’s written about his own struggle and has written on some of his friends as well.
There was a perfect storm of experience, seeing the downside of this market, we’ve seen what’s working over the past six years. Also, Josh had been the partner who led that investment course and worked with it. If there’s anyone I’m willing to bet on, it’s the person that I had previously bet my career on and started working with him as a Partner at Compound.
I do think it was kind of a culmination of like a ton of learnings internally as a firm, both on broader views within the healthcare ecosystem as well as in this more specific thesis we had been working on.
NA: Thanks so much for joining me, Michael!
MD: Thanks for reaching out!