Recurrence as a governance signal: diagnostic network metrics for public procurement oversight in Greece

This study examines what recurring buyer–supplier relationships in Greek public procurement reveal about routinised contracting behaviour. Drawing on contract award data from Greece's Central Electronic Registry of Public Contracts (KIMDIS, 2018–2025), it models procurement as a temporally evolving bipartite network across twelve domains, generating seven year-pair predictions. Three interpretable network signals — Historical Frequency, Preferential Attachment, and an adapted Adamic–Adar index — capture routinised continuity, structural concentration, and context-bound repetition. Predictability varies systematically across domains (AUC 0.80–0.96), with feature importance shifting from history-driven to structurally diverse recurrence over time. Authority-level analysis reveals extreme within-domain heterogeneity (vendor diversity 1–300+, HHI 0.008–1.0), demonstrating that uniform oversight thresholds are structurally inappropriate. The framework suggests differentiated governance responses — contestability reviews, dependency audits, and specification reform — and can be integrated into Greece's existing digital procurement infrastructure. Results are robust across three negative sampling specifications.