Against the backdrop of persistent global food insecurity and intensifying competition between humans and livestock for grains, enhancing the feed-use efficiency of agricultural by-products has become a critical pathway for ensuring feed security and advancing agricultural circular transitions. Using data from 57 counties in Gansu Province, China, this study integrates Data Envelopment Analysis (DEA) with fuzzy-set Qualitative Comparative Analysis (fsQCA) to systematically evaluate the feed-use efficiency of agricultural by-products in 2023 and to uncover its underlying configurational drivers. The results reveal that overall efficiency remains low and displays pronounced spatial heterogeneity, with insufficient scale efficiency emerging as the primary bottleneck. The fsQCA further identifies three high-efficiency configurations-namely the "technology-driven with scale support," "resource-abundant with scale leverage," and "productivity-oriented with scale foundation" pathways-reflecting a pattern of equifinality and conjunctural causation. These findings underscore the necessity of moving beyond uniform, one-size-fits-all strategies toward differentiated, context-specific approaches that promote coordinated interactions among technology, organizational capacity, and policy support. The study provides actionable insights not only for regions in China with similar resource constraints but also for dryland agricultural systems worldwide seeking to optimize resource utilization and accelerate green agricultural transitions.