Recent studies reveal inconsistent digital leadership-performance relationships in SMEs, challenging traditional direct-effect models that inadequately explain transformation mechanisms in resource-constrained contexts. This study resolves theoretical inconsistencies by testing a capability-mediated framework where digital capability fully mediates digital leadership-business performance relationships, with top management support as a moderator. Grounded in dynamic capabilities theory, we propose that leadership operates exclusively through capability-building mechanisms contingent upon organizational support conditions. Data from 210 Indonesian food and beverage SMEs were analyzed using partial least squares structural equation modeling with bootstrap procedures. The measurement model demonstrated excellent psychometric properties, while the common method bias assessment confirmed the validity. Findings reveal complete mediation, with digital leadership influencing performance exclusively through capability development, explaining substantial variance in organizational outcomes. Large effect sizes confirm digital capability as the primary performance driver and leadership as the key capability antecedent. Top management support significantly moderates the capability-performance relationship, where supportive contexts amplify returns from digital investments. Results advance digital transformation theory by establishing capability-mediated pathways as primary mechanisms, moving beyond traditional direct-effect models that showed inconsistent results across studies. For practitioners, SMEs should prioritize sequential capability-building strategies over simultaneous technology adoption, with leadership development preceding technology acquisition to maximize transformation effectiveness. This approach offers substantially higher performance returns compared to technology-first strategies. Cross-sectional design limitations necessitate longitudinal replication across industries and cultural contexts to strengthen causal inference.
Mendo et al. (Tue,) studied this question.