Home / Case Studies / Indonesian Fintech Fraud Detection
Case Study · Fintech · Fraud Detection
94% fraud detection accuracy
at enterprise scale.
The challenge
A fast-growing Indonesian fintech was facing an escalating fraud problem. Transaction volumes were growing faster than the manual review team could handle, creating both financial exposure and customer friction from false positives on legitimate transactions. A rules-based fraud engine was generating too many false negatives on novel attack patterns.
What PGI delivered
PGI-Data designed and deployed a real-time ML fraud detection pipeline on Google Cloud Platform, replacing the rules-based engine with an adaptive machine learning model capable of detecting novel fraud patterns at sub-second latency.
- Real-time transaction scoring pipeline on GCP
- ML model training on historical fraud data in BigQuery
- Feature engineering for behavioural anomaly detection
- Model serving via Vertex AI at sub-second latency
- False positive reduction through confidence scoring
- OJK-compliant audit trail and reporting
- Monitoring dashboard for fraud operations team
Measurable outcomes
94%
Fraud detection
accuracy
accuracy
<500ms
Transaction scoring
latency
latency
60%
Reduction in
false positives
false positives
Technology used
Working on a similar challenge?
Talk to our fraud detection specialists about your pipeline.