In the world of quantitative investing, new factors are proposed every day — some rooted in behavioral theory, others born purely from data mining. But the flood of ideas begs a harder question: what makes a factor “real”? At Astrai, we approach this question not just as statisticians, but as engineers and skeptics. We’ve developed a working framework to assess factor credibility using four dimensions: economic intuition, cross-market robustness, temporal persistence, and interaction with other signals. A good factor tells a story about investor behavior or market structure. A great factor works across regions and datasets. But a truly valuable factor holds up out-of-sample and in live trading — even when it’s no longer new. In this post, we walk through examples from our research — including factors we expected to work but didn’t, and those that surprised us. We also discuss how combining weak signals in intelligent ways can sometimes outperform chasing a single “strong” factor. Ultimately, we believe the bar for new factors should be high, but not impossibly so — and that thoughtful replication is the foundation of good innovation.