Alternative data for systematic investing

Navigating the data deluge

Over the last decade, the explosion of alternative datasets — from satellite imagery and job postings to shipping records and scraped web content — has radically expanded what’s measurable. Yet with this abundance comes a new challenge: discernment. Without a rigorous process, it’s easy to succumb to confirmation bias, leaning on a single dataset that ‘proves’ a hypothesis rather than generating genuinely additive insights.

As Figure 1 illustrates, the number of differentiated datasets available to our research team continues to grow across asset classes and geographies, reflecting both the increasing availability of alternative data and the escalating complexity in separating signals from noise. As illustrated in Figure 2, the inflow of datasets surged, yet our onboarding selectivity tightened — with the number of datasets rejected by our research team increasing fivefold from 2019 to 2024.

Distinguishing signals from noise

We evaluate prospective signals using a combination of quantitative testing, economic reasoning, and cross-signal validation. Our process focuses on seeking datasets that deliver predictive, economically intuitive signals that consistently forecast asset returns.

• Testing for statistical power: Frameworks such as Information Coefficient (IC), Predictive R-squared, and Horizon-Decayed IR help quantify dataset effectiveness.
• Economic intuition testing: Event studies, cross-sectional regression tests, or integrating the signal into a broader alpha model help validate rational investor behavior.
• Additivity and redundancy checks: Orthogonal testing against existing signals ensures differentiated, non-redundant insight.

Finding alpha potential in a sea of raw data

Sourcing, selecting, and scaling high-quality alternative data is no longer a niche capability — it’s a strategic imperative. In a world where signal libraries number in the thousands, finding the next source of alpha isn’t about being faster — it’s about being smarter.

We believe the future belongs to those who combine human insight, operational precision, and organizational scale. When every marginal dataset matters, how you source data becomes just as important as how you use it.