Policymakers are increasingly considering network effects when designing industrial policies, but the evidence is limited by data availability. Thanks to a unique combination of administrative records matched to the Estonian Value Added Tax data, which describe the quasi-universe of buyer-seller transactions in the country, this paper examines the structure of the Estonian firm-level production network. There are three key insights for industrial policy design. First, a few “superstar” firms are very central but also larger, older, more productive, more likely to export and import than most firms, underlying the network’s critical role in explaining aggregate productivity. Second, firm centrality increases with age, even when controlling for size, suggesting that industrial policies could target the network dimension to support firm growth. Third, this paper proposes a novel, data-driven approach for industrial policy, where communities of firms trading with each other would be the relevant unit of intervention instead of sectors.