The Future Analyst Is a Scientist, Not Just a Quant

Markets reward quick wins, but careers — and capital — demand the discipline of science.

Failing faster means enduring longer. Image generated with Leonardo AI

When I first started working with a major European bank on sustainability, I thought I had cracked the code. I built sophisticated correlation-based models. They looked impressive, and to a large extent they even produced the right numbers. I truly believed that my findings would help steer billions toward sustainable investments.

But there was a flaw: I wasn’t testing why things happened. The models were predicting, not explaining. They were good, but people rightly had their doubts about my claims. Whenever they asked, “well that’s cool but can you tell me why women would up the shareprice in that company or whether that’s just a statistically robust coincidence?”—in those moments (and there were a few of them) I stumbled.

That experience was humbling. It showed me how easy it is, even for someone with a scientific background, to stop thinking like a scientist once inside finance. To focus on prediction because it might make market alpha in the short term, while quietly abandoning the harder task of asking what really works, and why.

The truth is that, unless you’re a day trader, people do care about more than a descriptive trading strategy. Their human curiosity drives them to ask why things are happening. And that’s a good thing, because it helps design investment strategies that pay off much better in the long term.

Correlations come and go with the market’s ups and downs. But causal relationships endure like natural laws.

The analysts who endure, the ones who thrive over decades, are those who rediscover their scientific mindset. I’m not the only one who’s been there and learned it the hard way (even though I did it in my unique way, as an entrepreneur in sustainable finance, rather than as an employee in a leading financial institution).

So, to save others from making the same mistakes that I did, I’d like to ask: How do we bring that scientific mindset back into finance, at scale and also one individual at a time, before it’s too late?

The Limits of Prediction

Prediction is seductive. Correlation-based models can be built quickly, backtested beautifully, and presented with elegant charts. They make us feel smart. And in the short term, they sometimes even work.

That was my trap. Early on, I believed that if my models explained enough variance and generated sharp enough risk-adjusted returns, they would be self-justifying. The elegance of the mathematics became a substitute for real understanding. After all, who could argue with an R² that looked good on paper?

Luckily for me, I didn’t stumble over my own model falling apart — in fact, it has been holding up. It’s just that it didn’t sell, because I didn’t have a good answer to other people’s questions. “Why would women in management raise the share price? Couldn’t it just be a coincidence?” I rambled something about how the connection is so statistically likely that they should just go and make their money. But everyone in the room (including me) was dissatisfied with that.

Once that question — why — is on the table, the limits of prediction show fast. A model that can’t explain is not a strategy; it’s a gamble dressed up in numbers. That’s why whole fields — economics, policy, even tech giants like Microsoft and Netflix — have moved toward causal testing. Finance is only starting to catch up.

Science Adds Fast Failure Mechanisms

A scientific mindset flips the script. Instead of asking, “How well does X predict Y?” you ask, “Does X cause Y? And under what conditions?”

This shift sounds subtle, but it changes everything. Prediction gives you a surface-level pattern; causality gives you a mechanism. And mechanisms endure. When you know why something works, you can trust it beyond the next market cycle.

Whole disciplines have already made this turn. Development economists like Esther Duflo and Abhijit Banerjee built an entire Nobel-winning research program around randomized trials, precisely because correlations weren’t enough.

Econometricians like Angrist and Pischke have shown how quasi-experiments and identification strategies can pull real causal answers from messy data. And in finance, researchers such as Marcos López de Prado have been vocal about the need to separate causal effects from spurious correlations in machine learning.

For me, rediscovering this mindset was liberating. When clients asked “why,” I no longer had to stumble. I could point to mechanisms, to evidence, or — if the evidence wasn’t there yet — to a clear plan for how we might find it. That shift made my models stronger, but more importantly, it made me stronger.

In a world obsessed with short-term alpha, the real alpha is endurance — and endurance comes from science. Which is the world I came from!

Bringing Science Back (at Scale)

So how do we bring a scientific mindset back into finance — not just for one analyst at a time, but across entire organizations?

At the individual level, it starts with remembering where many of us came from. Most quants were trained as scientists: quant analysts are often former physicists, mathematicians, and engineers. We learned to form hypotheses, test them, and admit when they failed. Yet too often, that ethos gets stripped away once we enter finance. Rediscovering it means resisting the pressure to deliver quick answers and instead taking pride in asking the right questions.

At the firm level, it requires building structures that reward experimentation rather than just short-term performance. Tech companies like Microsoft and Booking.com run thousands of A/B tests each year; they’ve institutionalized the scientific method. Imagine if investment firms did the same: pilot new ESG-linked loan structures, test risk models in controlled rollouts, measure causal impact before scaling up. The result would be fewer costly mistakes and more robust strategies.

At the market level, the demand is already there. It’s not just the collaborators and (prospective) clients in my own bubble. Many long-term allocators — pension funds, sovereign wealth funds, insurers — are asking not just for performance, but for explanations. Regulators too are nudging in that direction, too, with stress tests and disclosure frameworks that implicitly demand causal reasoning.

In other words: finance doesn’t need to invent this mindset from scratch. The blueprint exists. The question is whether firms and analysts will embrace it before the market forces them to.

The Career Angle (Long-Term Survival for Quants)

There’s also a personal reason to care about this shift: your own survival as a quant.

Prediction-only quants often enjoy a good run. They find a clever signal, squeeze alpha from it, and for a few years they look brilliant.

But markets always shift. The signal decays, competitors catch on, and the model that once looked bulletproof turns into dead weight. Careers built only on prediction tend to flame brightly, then burn out fast.

Scientific quants play a different game. By asking causal questions, they build frameworks that don’t collapse with every new regime. They can adapt because they understand mechanisms, not just patterns. That makes them harder to automate away, harder to replace, and much more resilient over the long haul. But they also need more sophistication (not to mention more vulnerability when your new model still doesn’t work) to build up, as well as more time.

I’ve seen both sides up close. I know what it feels like to present models that impress but don’t satisfy. And I’ve felt the difference when I could answer the tough “why” questions with confidence. That confidence isn’t just good for clients; it’s good for your own career.

If you want to be more than a one-hit wonder in finance, the best hedge you can buy is scientific curiosity. It compounds. It pays dividends. And it keeps you in the game long after the short-term predictors have been sidelined.

The Bottom Line: Scientific Minds Win the Long Game

Finance has always rewarded speed. The fastest model, the sharpest backtest, the cleverest signal — these win applause in the moment. But applause fades quickly. The market forgets last quarter’s winners, and even brilliant predictors eventually face the same fate: their models stop working.

What endures is explanation. If you can show exactly why something works, you create value that lasts beyond a single market regime. That’s the essence of science, and it’s what finance needs most right now.

For capital allocators, this means shifting the focus from shiny signals to durable strategies. A predictive model might steer money well for a few years, but a causal insight can shape allocation for decades. That’s what pension funds, insurers, and long-term investors are really looking for: not just “what worked,” but “what will keep working, and why.”

For analysts and quants, the stakes are just as personal. Prediction-only careers burn bright and burn out. Scientific careers evolve, adapt, and compound. It’s true that they look less-than-spectacular in the beginning and still feel like a stony road midway (I’m speaking from experience). But—the more you embrace curiosity, hypothesis testing, and causal reasoning, the more resilient you become, both to market shifts, and to the natural turnover of quant fashions.

The market doesn’t need more predictors. It needs more scientists. And the quants who remember their scientific roots will be the ones still standing — not just at the end of this cycle, but at the end of their careers.

---

Reads of the Week

• This article from LLMQuant explores how new tools are making large language models (LLMs) more transparent and accountable in high-stakes finance. It introduces mechanistic interpretability—a method for reverse-engineering neural networks to understand how they make decisions at the neuron level. Unlike traditional "explainable AI," which offers surface-level insights, this approach digs into how models process concepts like credit risk or fraud, allowing for bias detection, audit-ready explanations, and safer deployment in areas like credit scoring, trading, and compliance.

• For our quant-savvy readers, here’s a helpful refresher: this piece from AI-Driven Quant Investment Strategies demystifies Alpha and Beta by anchoring them to a simple linear regression line. Beta (slope) reflects a stock’s sensitivity to market moves—think sector-wide behavior—while Alpha (intercept) captures the return that’s independent of the market, a.k.a. stock-specific performance. It’s a clean, visual explanation that’s especially useful if you're building or tuning AI-driven factor models, and a timely reminder that some of the most powerful financial insights are grounded in elegant simplicity.

• In this edition of Agentic Equities, we get a deeper look into how ChatGPT’s stock picks respond to real market movements—and the answer is: unpredictably. While GPT price targets shift about as much as market prices week to week, they don’t correlate strongly with actual price changes, likely due to the randomness built into how LLMs generate responses. For quants, this opens up a fascinating new angle: ChatGPT’s market impact might be less about consensus and more about distributed volatility—one more variable to consider in an increasingly agentic market.