This is a sponsored advertorial from Robeco.

Small caps offer distinct return drivers and remain under‑researched. Next‑gen quant is starting to unlock their potential.

Small caps have very different return drivers from large caps, making them a powerful diversifier. Yet structurally, small caps remain one of the least efficiently researched parts of the global equity markets. 

The NextGen evolution: What machine learning brings

With more than 4,000 stocks worldwide and significantly lower analyst coverage than large caps, the universe combines breadth with informational inefficiency – a powerful combination for quant investors. Quant strategies process large datasets, apply consistent definitions, and update signals with discipline, turning the small-cap ‘needle-in-a-haystack’ problem into a repeatable selection process.

As Mike Chen, head of NextGen Quant at Robeco, put it, “As small caps are a very rich and idiosyncratic hunting ground for quant, they are also the perfect environment for machine learning to shine.”

Robeco has recently launched a new global small-cap equities strategy that builds on decades of quant research using our range of established and next-generation quant signals as input. The strategy then takes a step further by selectively applying machine learning and alternative data to capture more complex market dynamics.

Machine learning helps in two key ways:

• Identifying non-linear relationships, where signals behave differently depending on context
• Capturing interaction effects, where combinations of signals matter more than each individually

“We let machine learning determine how features should be combined for each stock,” Chen explained. “Some companies are driven more by value, others by quality. The model adapts dynamically.”

This flexibility matters particularly in small caps, where company drivers are more diverse and less correlated. “Machine learning allows us to extract nonlinear, higher-order dynamics that traditional linear models simply cannot capture,” he added.

Keeping the guardrails: Explainable, benchmark-aware implementation

While AI enhances stock selection, implementation discipline remains central. The strategy is constructed within clear constraints across sectors, countries, and regions, with controlled turnover and governance oversight. It is designed to remain benchmark-aware while seeking excess returns.

Crucially, the model is fully interpretable. “We don’t see this as a black box. It’s a glass box,” said Chen. “We have tools to attribute performance and understand exactly which signals are driving portfolio decisions.” This transparency extends across both stock-level positioning and portfolio-level outcomes, ensuring the strategy remains explainable for investors and risk teams alike.

Human oversight in an AI-driven process

Even the most advanced models operate within the limits of their assumptions, which is why human oversight remains integral. “Models are simplifications of reality,” Chen noted. “Portfolio managers monitor assumptions constantly – especially during extreme events like Covid, where market regimes shift suddenly.”

Oversight does not mean interfering with day-to-day model outputs. Instead, it focuses on validating whether the structural assumptions underpinning the model remain intact. This balance preserves the strengths of systematic investing, such as discipline, scalability, and bias reduction, while ensuring resilience during periods of structural change.

The investment case for small-cap equities

While small caps have outperformed large caps over the last 20+ years (see Figure 1), they have been lagging since 2018. However, Figure 1 also reveals that changes in relative valuations between these segments were the main reason for this underperformance.

Figure 1: Relative performance and valuation of small vs large caps in developed markets

Historically, small caps have traded at valuation premiums of up to 30% versus large caps. Today, they trade at a 30% discount – a gap not seen over the last 20 years. From an allocation perspective, this creates both cyclical and structural appeal.

ETF implementation: Innovation that’s usable

The strategy is delivered through an active ETF wrapper combining a research-driven stock-selection engine with intraday liquidity, transparency, and operational efficiency. From a portfolio construction perspective, the ETF is designed as a diversifying small-cap allocation with controlled active risk.

With a tracking error of roughly 3-4%, it sits between enhanced indexing and more thematic exposures on the active risk spectrum, offering structured alpha potential without concentrated bets.

The next evolution, not just the next label

In recent years, many strategies have adopted an AI label. The more important question for investors is whether the investment process meaningfully evolves. In other words, does it improve outcomes without sacrificing discipline, transparency, or risk control?

For Robeco, the NextGen Quant platform is designed to do precisely that. “NextGen is about pushing the frontier,” Chen said. “We’re continuing to build the suite, launching new strategies while constantly iterating and improving existing ones.”

Robeco’s NextGen Global Small-Cap Equities strategy reflects that philosophy: combining an established quant heritage with modern AI techniques, applied where market structure makes them most relevant.

This is a sponsored advertorial from Robeco.