<p>Vast sales data in the retail sector present opportunities to optimize operations and increase revenue. Consequently, studies have leveraged these data to predict sales, detect anomalies, segment customers, and similar. Anomaly detection techniques have been proposed to identify deviations in sales patterns at both the product and product-store levels. However, these studies focus on a single store or a few stores without considering sales patterns across similar stores, which could greatly benefit retailers. For instance, if a store outperforms similar stores in sales of a specific product, best practices could be established and shared throughout the retail network, ultimately increasing overall sales. Nevertheless, cross-store anomaly detection is challenging due to contextual differences among stores, diversity of sales patterns, and the variety of products. Consequently, this paper proposes an unsupervised collective-contextual anomaly detection framework for the identification of outperforming stores in massive sales data from a large retail network. In contrast to other studies, which consider intra-store patterns, we integrate intra- and inter-store anomaly detection. The proposed approach addresses the gap by decoupling the problem into two components. Collective anomalies are detected by leveraging a deep-vanilla autoencoder to identify patterns that deviate from normal behavior. Then, contextual anomaly detection leverages the designed product-level similarity metric, to identify similar stores and examine patterns among them. Experiments using a dataset from a major Canadian retailer demonstrate the success of the proposed approach, achieving 90% overall precision, with collective anomaly detection outperforming other conventional approaches by achieving 85% precision, 72% recall, and 79% F1-score.</p>