The exponential growth and variety of studies on returns highlights the need for a comprehensive bridge between theory and empirics. The paper addresses this demand by proposing a hierarchi- cal state-based asset pricing model built on two interconnected solutions: the state space of assets and explanatory gain decomposition approach. Inspired by the reinforcement learning applications, the state space of assets represents a generic and consolidated description of the stochastic environment based on observable fundamental and macroeconomic characteristics. The decomposition approach in turn tackles the complexity and heterogeneity of underlying effects enabling a switch from kitchen-sink opaque analyses to a hierarchy of piecewise transparent models. The proposed model is demonstrated along the path from asset valuation to multifactor applications inclusive of boosting the explanatory power of price-to-fundamental regressions, resolving the weak correlation between the HML and RMW factors, and shedding light on HML performance through cross-sector dynamics. Overall, the state-based approach constitutes an alternative and complementary direction to multifactor models, extending the focus of empirical research to under the hood of factors with a hierarchical state space of assets.