AbstractA large part of biodiversity theory is driven by the basic question of what allows species to coexist in spite of a confined number of niches. A substantial theoretical background to this question is provided by modern coexistence theory (MCT), which rests on mathematical approaches of invasion analysis to categorize underlying mechanisms into factors that reduce either niche overlap (stabilizing mechanisms) or the average fitness differences of species (equalizing mechanisms). While MCT has inspired biodiversity theory in the search for these underlying mechanisms, we feel that the strong focus on coexistence causes a bias toward the most abundant species and neglects the plethora of species that are less abundant and often show high local turnover. Given the more stochastic nature of their occurrence, we advocate a complementary cross‐level approach that links individuals, small populations, and communities and explicitly takes into account (1) a more complete inclusion of environmental and demographic stochasticity affecting small populations, (2) intraspecific trait variation and behavioral plasticity, and (3) local heterogeneities, interactions, and feedbacks. Focusing on mechanisms that drive the temporary coviability of species rather than infinite coexistence, we suggest a new approach that could be dubbed coviability analysis (CVA). From a modeling perspective, CVA builds on the merged approaches of individual‐based modeling and population viability analysis but extends them to the community level. From an empirical viewpoint, CVA calls for a stronger integration of spatiotemporal data on variability and noise, changing drivers, and interactions at the level of individuals. The resulting large volumes of data from multiple sources could be strongly supported by novel techniques tailored to the discovery of complex patterns in high‐dimensional data. By complementing MCT through a stronger focus on the coviability of less common species, this approach can help make modern biodiversity theory more comprehensive, predictive, and relevant for applications.