Automated Monitoring of Exception Scenarios in Store Operations Using the BCP System

QA traditionally focuses on preventing exceptions through risk assessment and testing, but automated monitoring of exception data adds a defensive layer that enables rapid response when anomalies occur, forming a closed‑loop with prevention.

The current problem‑discovery methods—store staff feedback, online issue reports, business alarm codes, and log analysis—are slow and can cause significant financial loss, especially for coupon and red‑packet distribution or product status inconsistencies.

To address timely detection, the store integrates the BCP system, which decouples business validation from code by listening to MQ messages and database binlog changes. The system supports two rule‑engine modes: Aviator (no code needed) and Java (extends AbstractJavaRuleExecuteEngine with when and then methods).

Example rule for validating the consistency of a recycle order’s pickup status is shown, with the when part extracting message fields and performing simple null/value checks, and the then part establishing a data source connection to execute SQL validation.

Before integration, common problematic scenarios were identified, including concurrent product status errors, coupon issuance anomalies, mismatched recycle order statuses, and price inconsistencies between store and product sides.

After deploying 20 validation rules (7 for recycle, 11 for retail, 2 for infrastructure), three main validation approaches are used: (1) Aviator rule engine with filter conditions and data checks, (2) Java rule engine with custom logic calling RPC interfaces, and (3) Binlog monitoring with delayed re‑query of database tables.

During the monitoring period, 671 blocked‑order records were discovered (6 retail, 664 recycle). Retail anomalies included concurrent status issues, inventory mismatches, and price discrepancies caused by store‑initiated price changes not reflected on the product side. Recycle anomalies were mainly blocked orders, which required manual verification before reporting to operations.

To reduce manual effort, the team plans to automate the aggregation of BCP‑generated alerts, configure statistical rules, and close the loop by marking resolved alerts after operations handle them, thereby improving efficiency and accuracy.

The significance of this work includes improved product flow and reduced inventory backlog, deeper QA understanding of business logic leading to faster issue triage, and early detection of hidden technical problems, ultimately establishing a robust exception defense wall.