<![CDATA[Digital transformation has reshaped partner evaluation in agribusiness business-to-business (B2B) networks, shifting decision-making from intuition-based judgments to transparent, data-driven assessments. Addressing the need for scalable and trustworthy selection mechanisms, this study introduces a novel hybrid anomaly detection framework that sequentially combines rule-based z-score normalization with the Local Outlier Factor (LOF) algorithm to evaluate digital business credibility. The framework leverages Google Maps data, a widely accessible, user-generated information source that reflects real customer experiences, to assess 6,237 hospitality, restaurant, and café (HORECA) businesses in Indonesia’s Jabodetabek region, a growing hub in the agribusiness supply chain. Using structured data collected through the Google Places API, the rule-based method identified 47.06% of businesses as anomalies, predominantly those with disproportionately high ratings relative to customer engagement. Meanwhile, LOF detected 5.02% of density-based outliers, capturing irregularities that only emerge in local spatial and contextual comparisons. A statistical comparison (χ² = 195.10, p < 0.001) revealed a 56.52% overlap between the two methods, emphasizing their complementary strengths: rule-based thresholds provide interpretability and efficiency, whereas LOF offers sensitivity to nuanced, neighborhood-level deviations. These findings show that no single technique fully captures the complexity of digital credibility anomalies; however, their combination enables more balanced and context-aware evaluations. This approach enhances the accuracy and fairness of credibility assessments, which is crucial for partner selection in digital agribusiness ecosystems. Overall, the study provides practical and methodological contributions for building transparent, reproducible, and equitable anomaly-detection systems for emerging digital markets]]>
