Sink or Swim: How to Combine Quant and Traditional Asset Management Techniques
What this covers
This is a conversation between Leigh Drogen, founder of Estimize, and host Patrick, working through what a discretionary fund would have to build to use quant tools. Drogen's blueprint is concrete. Analysts submit forward-looking estimates for EPS, revenue, EBITDA, and company-specific KPIs such as Netflix subscriber growth into a central database several times a quarter. Fuzzy judgments like governance quality become numeric ratings, so they can be tested later. Machines handle beta-neutral portfolio construction, while the PM decides macro exposure, factor tilts, and how to size outlier calls. A Virtual Portfolio Benchmark built from each analyst's top picks shows whether the PM or the analyst is adding value.
The range goes well beyond mechanics. Drogen draws on Clausewitz, Sun Tzu, and the US in Afghanistan to argue that tactics without a political goal fail. He argues that consensus ideas can't produce alpha, that firms reward lucky outcomes instead of decision quality, and that Union Square Ventures' worst investments were the ones every partner liked. A Microsoft-versus-Tesla comparison separates Beta from Alpha, and he cites Estimize data showing no correlation in forecasting accuracy across sectors. One contested point: he doesn't reject beta entirely, and describes taking a macro view from breadth indicators, as his mentor David Geller did.
Patrick pushes toward Simple Systems Over Complex Ones, such as force-ranking analysts and equal weighting, and offers a definition of alpha as patience and idiosyncratic risk uncorrelated with cheap Smart Beta. The talk closes on hiring: Drogen says the scarce candidate knows both finance and data, and can write.
Drogen argues traditional discretionary hedge funds can only compete with quants by restructuring around systematized, structured analyst inputs and a PM who acts as coach and capital allocator, not stock picker, because ego and unstructured politics currently block alpha measurement.
- Ego and PM control over decisions, not tactics, is the binding constraint on transition
- Structured, forward-looking analyst estimates let firms measure who adds alpha, which unstructured reports prevent
- Simple systems beat complex ones, and specialization and measurement matter more than sophistication
Quant adoption fails unless the PM's ego is subordinated to a political decision structure.
- The PM's ego and desire for control over decision-making and power cause ideas to funnel up to the PM, when the PM should instead act as offensive coordinator, calling plays for a team that generates ideas below them.
“The political aspect is mostly the ego associated with PMs having control over the investment process and final decision-making”
- Hedge funds trying to adopt quant tools are doing it backwards: without a political goal, meaning a clear decision structure in which the PM's ego is subordinated to process, adopting tactics and strategies will not produce a winning outcome, just as the US in Afghanistan used right tactics without a political objective.
“if you don't have a political goal, it doesn't matter what tactics and strategies you use to fight the battle, you're going to lose in the end”
- The easiest response for funds that woke up to quant is to hire five data scientists and an engineer, buy unique data, and throw it at the PM, but this fails because the PM will not engage with a process from a different way of thinking.
“This simply doesn't work because yeah, these are two different worlds, two different ways of thinking. The PM won't pay attention to it.”
- Some firms tried making quant teams operate like sell-side shops charging PMs through P&L, but this fails because the arrangement avoids the political work of integrating the two groups, and those firms are now retrenching and cutting the cost center.
“There are some firms that have tried to do this on a larger scale, and get the PMs to actually pay for it like through PnL”
- The CIO has to be the firm's strongest backbone because PMs effectively are the firm today, and the CIO must spend political capital to put a rigorous measured process in place, even if it upsets PMs.
“the CIO, at this point, has to be the one with the strongest backbone because they're going to have to stand up to the PMs.”
- Drogen claims women should dominate the PM coaching part of the process because they have less ego and less irrational behavior in egotistical professional environments.
“frankly, I have to say that women probably should rule this part of the process. They're simply better at this.”
- Investors should know the final outcome they want before acting, avoiding fighting a war on a tactic like terrorism, because a clear objective defines what to do, and Sun Tzu's lesson is to know you have won before engaging.
“when you get involved in anything, know what the final outcome you want is. Don't just say like, 'We're going to go fight a war against terrorism.' That's a war on a tactic.”
Analysts must log forward estimates and unstructured judgments as structured numeric data.
- Analysts should be required to submit forward-looking estimates for EPS, revenue, EBITDA, and company-specific KPIs such as Netflix subscriber growth or Starbucks same-store sales into a centralized database, refreshed about three times per quarter, so the firm can see and track each analyst's views.
“analysts have to give a centralized database of forward-looking view on EPS revenue, EBITDA, and then whatever key performance indicators the company is really kind of bound with”
- No firm fully runs the platonic integrated process of fundamental and quant; firms have pieces, such as analyst estimate tools and Omega Point's portfolio optimizer, but nobody has built central software that runs a weekly review of ranked names, factor scores, and analyst accuracy.
“nobody's really build a piece of software either internally or externally, from these funds that actually says, 'You're going to review this on a Monday meeting every week.'”
- Firms claiming to be long-term investors often trade out in a quarter because they never wrote down the structured variables behind a long-term thesis, so they cut ideas before they play out.
“if you don't have all the structured variables that add up to why you believe in a long-term thesis, you will cut and run much earlier”
- Funds that have quants run semantic analysis on the free text analysts enter into tools like Tamale and CodeRED are trying hard to do the wrong thing, because the text needs to be structured before any quant can use it.
“You're putting so much effort into the wrong thing, and it's really funny.”
- Unstructured analyst judgments such as corporate governance or probability of a buyout should be recorded as numeric fields on a fixed scale, such as a one-to-ten rating each quarter, so the firm can later test whether those ratings correlate with volatility, earnings misses, or stock performance.
“my field is corporate governance, and each time I go in there to like update my whatever, I have to rate corporate governance for this company on a one to ten scale”
- When Estimize pitches its data, asking firms how they would use it exposes that many do not have a defined process, since analysts bring ideas and the PM discusses them ad hoc with no formal meeting.
“They're like, 'Well, the analyst brings the ideas. Then we discuss it.' ... 'No, not really. It happens ad hoc.'”
PM value should be measured as alpha from factor and market timing, not beta.
- Over six months, Microsoft's gains were mostly beta, since they could have been captured with a smart-beta ETF, while Tesla's gains were almost entirely alpha because it was uncorrelated with the market and momentum factors.
“If you invested in Microsoft over the last, say six months, you've created a positive return, but almost all of that positive return is beta.”
- A firm can measure whether a PM adds value by building an equal-weighted virtual portfolio of each analyst's top five long and top five short picks and checking whether the PM's actual portfolio beats it; if not, the PM should be replaced by the model, and if the analyst is bad, the analyst should be replaced.
“I can judge the alpha the PM is creating by how much they deviated from those picks”
- Every year firms should measure how much alpha PMs created by toggling factor and market exposure, and fire them or keep them based on the result, because the PM's macro timing is supposed to be the value they add.
“every year, we're going to look at that amount of alpha they created by toggling things up and down, and they're either going to get fired or they're not”
- Large outlier PM bets are a small sample, which makes it hard to evaluate PMs on them, but if PMs are supposed to add value through such bets, firms must either hold them accountable with measurement or drop the PM role entirely.
“At this point, you either hold the PMs feet to the fire, and measure like that, or just don't have a PM, just have a quant on it.”
- Portfolio construction should be done mostly by machines that beta-neutralize positions, while the PM adds value by deciding how much macro exposure to take at different times, tilting factor exposure such as momentum, and sizing outlier analyst picks.
“Most of the construction should be done by a machine at this point, these systems.”
- Drogen describes his own momentum philosophy as buying high and selling higher, using relative strength, and exiting before margins collapse at the top, riding the multiple up rather than catching it at the bottom.
“You want to buy high and sell higher. You want to use relative strength in your favor, and then you want to get the hell out of there before margins collapse at the top”
Analyst track records by sector should be measured and used to weight their views.
- Analyst track records by sector can be turned into confidence intervals displayed beside each pick, so a PM can challenge the largest upside call when the analyst's historical accuracy in that setting is low and require more evidence before sizing it.
“we can put really easy confidence intervals next to their picks on that dashboard”
- Estimize's data shows analyst skill follows a normal distribution, with weaker forecasters leaving faster, and skill is persistent enough that Drogen says the data environment is less random than the team expected, with luck separable from skill.
“We were not surprised to find that it is a normal distribution of skill”
- Militaries promote generals after lucky outlier victories even when their bets were bad, and finance does the same by rewarding lucky outcomes, so firms should measure decision quality rather than outcomes, since a lucky run of heads is followed by random results that look like poor performance.
“This is why I think measurement versus outcome is so important because you might flip heads five times in a row”
- In Estimize's data, analysts who made between 10 and 50 company estimates per quarter were more accurate than those who made fewer than 10 or more than 50, because fewer suggests lack of engagement and more suggests generalism and thin conviction.
“we find that analysts who made more than 10 estimates, but less than 50 per quarter... were more accurate”
- In Estimize's data, forecasting accuracy showed zero correlation across sectors, so analyst weights should be based on sector-specific track records and not on general accuracy outside that sector.
“we found zero correlation between accuracy between one sector and another.”
Traditional hierarchy hides accurate, contrarian analyst ideas, so analyst skill goes unjudged.
- In the traditional hierarchy, a good analyst is not necessarily a good politician inside the firm, so analysts with accurate, differentiated ideas often fail to sell them, and PMs then ignore those ideas, making analyst accuracy impossible to judge.
“The issue is that a good analyst isn't necessarily a good politician within a firm”
- Ideas that everyone finds obviously sensible are usually what gets put into portfolios, while controversial ideas get thrown out, yet by definition a consensus position cannot generate alpha, so PMs systematically select the wrong ideas.
“ideas that are like, 'Oh yeah, that makes total sense,' are usually the ones that get put into the portfolio”
- Union Square Ventures found its worst investments were the ones where partners all agreed the idea made total sense, so a good investment idea should draw a barbell reaction of enthusiasm and disbelief.
“They find that their worst investments are the ones where everybody is like, 'Oh yeah, that makes total sense.'”
- Fred Wilson says he knows he has made a great investment when other people call him to ask why he did it, which Drogen frames as evidence that great investments look contrarian at the time.
“Fred Wilson will say he knows he's made a great investment when other people call him up and say, 'Why the hell did you do that?'”
- Subjective, forward-looking expectations that differ from the rest of the market are the core of generating alpha, so the fact that inputs are subjective is acceptable and even desirable.
“That is the core of generating alpha is the subjective nature of your forward-looking expectations being differentiated from the rest of the world.”
- Behavioral economics teaches that markets are irrational because humans are irrational, and that specific patterns of that irrationality can be identified and exploited by investors.
“in behavioral economics, you learn that markets are incredibly irrational because humans are irrational”
Firms must write down core investment beliefs and style before building any process.
- A firm should codify its stated style and core beliefs in writing, like a constitution, because managers who drift from their stated style underperform, and trades outside the strategy become the largest losers since the firm lacks risk management for them.
“Every firm should have a stated style, and we know that when you have a specific strategy, and you try and take trades outside of that strategy, those trades are going to be your biggest like losers.”
- Many firms do not know their core philosophical beliefs about market cap, sector, growth versus value, and momentum, and the rest of the process cannot be built correctly until those beliefs are specified because every later step is aligned to them.
“I think the problem is a lot of these firms don't even know what their core philosophical beliefs are, and you can't build the rest of the process without that”
- Tearing a fund down to the studs means rebuilding from core beliefs rather than bolting new tools onto existing structure, which is the first step in both building or rebuilding a firm's process.
“could you please actually rip our whole process down to the studs?”
- Drogen predicts that only about 20 to 30 percent of current hedge funds will survive the shift to structured, systematized processes, and that people from failing firms will found new firms built around the process from scratch.
“I really honestly think it's only 20 to 30% of them are going to survive the shift”
Analyst estimates should be compared against market expectations to find where PMs must challenge.
- The PM still needs to exist because software is poor at spotting outliers; the PM should focus on the biggest positive and negative deltas between analyst and market expectations, challenge the analyst with specific questions, and act as coach rather than overruling the analyst.
“the core skillset of the PM has to go from being a stock-picker to basically being a coach.”
- Quants should work with PMs and data analysts to identify what a sector's names share, then turn datasets into factor models that score every stock, such as ranking historical earnings-surprise tendency in technology, which is correlated with future momentum and outperformance.
“history surprise of analyst estimates is a factor that is correlated to future momentum and out-performance in stocks”
- Fundamental investing can be expressed as a forward one-year estimate multiplied by the expected future multiple, such as Apple's EPS times a multiple rising from 15x to 18x, and comparing this price target with the market's consensus-implied target reveals where the analyst disagrees most.
“You multiply that multiple by your forward-looking estimate. That's your price target.”
Firms need hybrid hires who understand finance and data, with analysts specialized by sector.
- The biggest supply-demand imbalance is that asset management has too many economics and political science graduates and is hiring pure data scientists who know no finance, so the valuable candidate is someone who knows both finance and data and can also write and communicate.
“Right now, there's just too many Econ and Poli Sci people. Then, they're trying to hire all these pure data science people that know nothing about finance”
- When building a quant function, firms should hire data engineers first because they feed data to analysts, and then hire data analysts, a hard-to-find hybrid who understands stock drivers and models and can also manipulate data.
“you need to start with the data engineers because they're actually the first ones that are going to be flowing data into the analysts”
- Analysts should be deeply specialized in one sector and ideally come from industry, so their information is differentiated, while generalism belongs to the PM and CIO level that allocates across sector books.
“I think the analyst needs to be super specialized.”
The systematic approach suits bounded horizons and fails on rare special situations and complexity.
- The quant-plus-fundamental framework described applies to horizons from weeks to about two or three years, and differentiated views on culture or long-run industry trends would need different structured variables and factor models.
“We're talking weeks to two years, three years.”
- Systematic processes miss one-time special situations, which are essentially bets that rely on rare big ideas; firms that are not Soros or Buffett lack the capital to exploit them as a structural advantage.
“You're not Soros. I'm not Soros, and people out there aren't going to be Soros. We're also not Buffet.”
- Financial people love complex things because they believe complexity equals quality, and this is partly a marketing tool; only a handful of funds such as Renaissance Technologies truly benefit from complexity, and the rest should avoid it.
“Financial people love complex things because they believe if it's more complex, it must be better. It's just simply not true.”