Risk Management System (RMS) Data Warehouse: Real-Time Early Loss Detection Across 240 Nationwide Units

Operating across 240 nationwide entities—comprising head office divisions, regional branches, and operating units—operational and financial risk surveillance presents an ongoing challenge for major commercial banks. The modernization of the Risk Management System (RMS) Data Warehouse, engineered by Evotek, was driven by a direct, vital objective: empower the bank to detect early indicators of operational loss in real time, moving beyond retrospective monthly ledger reviews.

Structural Friction in Wide-Scale Banking Operations

Prior to the modernization, the banking institution faced operational hurdles inherent to a large-scale enterprise network:

• General Ledger (GL) losses were discovered too late: Operational breakdowns and unauthorized activities typically reflect as aberrant postings in ERP accounting ledgers. However, periodic manual audits depended on month-end financial closing, delaying risk reports to the Head Office risk division by weeks and forfeiting critical windows for preventative action.
• High labor overhead in bilateral incident reporting: Gathering loss incident files, reconciling dispute minutes, and exchanging physical dispatches between 240 branches and the Head Office consumed thousands of expert man-days annually.
• Gaps in non-financial risk oversight: The legacy platform focused solely on traditional operational risk, lacking specialized tracking for IT Risk (cybersecurity threats, core digital service outages), Reputational Risk (media escalations), and Strategic Risk.

Architectural Blueprint: Real-Time Stream Processing with Kafka and ClickHouse

To concurrently master high-throughput ingestion and sub-second analytics across 240 units, Evotek deployed a high-performance modern data architecture:

1. Real-time ERP ledger change capture (Kafka Streaming): We integrated directly with the bank’s ERP accounting core using Change Data Capture (CDC). As soon as accounting entries post to sensitive GL accounts, transaction streams are fed into Kafka clusters for instantaneous processing.
2. Automated anomaly detection engine: Implementing rolling statistical algorithms (Z-score 3-sigma over 90-day historical turnover) on individual GL accounts, the engine automatically flags abnormal transaction spikes or rule discrepancies, instantly triggering alert tickets to Head Office and branch risk officers in seconds.
3. High-throughput ClickHouse analytical warehouse: To prevent heavy analytical aggregations from congesting transactional databases, risk data is routed into a ClickHouse columnar database, generating real-time risk heatmaps and multidimensional reports across 240 branches instantaneously.
4. Unified non-financial risk modules: Standardized independent workflows for IT Risk, Reputational Risk, and Strategic Risk were introduced, assembling a comprehensive 360-degree risk management landscape.

Measurable Business & Operational Impact

Deploying cutting-edge data engineering directly into live daily operations delivered marked enterprise outcomes:

• 2,040 expert man-days saved annually: Automating collection, verification, and escalation workflows between the Head Office and 240 branches reduced administrative manual labor, yielding approximately VND 3.95 billion in annual operational savings.
• Near-instantaneous loss detection: Rather than waiting 15 to 30 days after month-end book closes, abnormal ledger patterns are detected within seconds, allowing risk officers to safeguard capital and prevent asset leakage proactively.
• Auditable, digitized collaboration: An interactive digital portal connects Head Office with branches, ensuring every risk ticket has a designated assignee, clear remediation SLA, and complete audit trail compliant with State Bank of Vietnam Circular 13.

Today, RMS functions not as a passive data repository, but as a digital watchdog empowering the banking institution to scale operations safely upon a resilient, modern foundation.