The Client
One of India’s largest non-banking financial companies, this organization operates across personal loans, vehicle finance, home loans, and SME credit products. Managing a large and growing book across multiple Lines of Business, it relies on timely, accurate data flowing between its CRM, Loan Origination System, Loan Management System, and Collections platforms to drive credit decisions, regulatory reporting, and customer service.
The Challenge
In a market where speed is a competitive differentiator, relying on legacy batch architectures leads to critical revenue leakage and customer churn. These technical limitations prevent the organization from achieving the real-time data visibility required for prompt credit decisions.
- Opportunity Loss from Delayed Data: Loan sanction decisions required data from CRM, CIBIL, and LOS. Batch delays of 1–2 days caused missed opportunities and slower credit decisions in a market where speed is a competitive differentiator.
- Data Inconsistency Across Systems: The same loan records were updated at different times across CRM, LOS, LMS, and Collections, creating conflicting data views and degrading customer service quality.
- Inaccurate and Out-of-Order Reports: Transactions arriving late to data marts produced reports out of sync with the actual business state, eroding trust in the enterprise data warehouse.
- Slow LOB Onboarding: Integrating a new Line of Business into the data platform took weeks to months, limiting speed-to-insight for new product launches and acquisitions.
- Scalability Ceiling: The legacy batch architecture lacked the flexibility to scale with a growing source estate or accommodate real-time analytics demands.
The Impact
- 33 Enterprise apps onboarded
- 24 hours New LOB go-live, down from weeks
- 4 months End-to-end delivery
The Solution
Hoonartek implemented an enterprise CDC platform on GCP built on open-source Debezium and Apache Kafka, replacing the client’s fragile batch ETL pipelines with near-real-time data propagation from all major source systems into BigQuery. The platform supports Oracle, MSSQL, MySQL, Postgres, MongoDB, Salesforce, and GCS file-based sources under a single ingestion framework, covering the full enterprise source estate in one architecture.
To accelerate deployment and reduce operational overhead, Hoonartek built a suite of purpose-built accelerators: automated BigQuery dataset and table creation from source metadata, template-driven connector generation for both Debezium source and Kafka-to-BigQuery target connectors, and a three-step onboarding utility enabling any new LOB to go live in real-time sync within 24 hours. Schema drift detection, automated reconciliation, and Secret Manager integration for secure credential handling were built in from the outset, delivering audit-grade data integrity across all 33 onboarded enterprise applications. The entire platform was delivered in 4 months with zero proprietary licensing cost.
Key Benefits
- Near-Real-Time Data Availability: Loan, collection, and customer data now propagates from source systems to BigQuery in near real time, reducing reporting lag from 1–2 days to minutes.
- Rapid LOB Onboarding: A three-step utility workflow enables a new Line of Business to go live in real-time sync within 24 hours, compared to weeks with the legacy integration approach.
- Zero Proprietary Licensing: Built entirely on open-source Debezium and Apache Kafka, no vendor lock-in, no proprietary software licensing; the client pays only for the managed service.
- Audit-Grade Reconciliation: Daily automated record-count verification and metadata reconciliation across all sources ensures data integrity and supports regulatory audit requirements.
- Eliminated Brittle ETL: Debezium-based CDC replaces hand-built batch scripts, with schema drift detected and handled automatically, removing ongoing manual pipeline maintenance.