How CreditVigil Works
A technical deep-dive for risk teams evaluating independent model monitoring.
Connect Your Data — Three Ways
Whether you prefer manual uploads or real-time database connections, we meet you where you are.
CSV Upload
Best for Lite plan users
Upload monthly loan-tape snapshots through our secure portal. We validate the schema, flag anomalies, and process your data within minutes. No engineering effort required.
API Integration
Best for Pro plan
Automated data feeds via our REST API or SFTP. Schedule daily or weekly pushes from your core banking system. We handle retries, deduplication, and versioning.
Direct DB Connection
Best for Enterprise
Real-time streaming from your database via read-only replicas. We connect to PostgreSQL, MySQL, or MongoDB and pull fresh data on your schedule. Zero manual effort after setup.
Sample CSV Schema
loan_id, origination_date, score_at_origination, loan_amount, term_months, product_type, current_status, days_past_due, snapshot_date LN-001, 2024-01-15, 720, 50000, 12, personal, current, 0, 2025-02-01 LN-002, 2024-02-20, 580, 25000, 6, sme, default, 90, 2025-02-01
What We Monitor
Four families of metrics that cover population drift, model discrimination power, and portfolio health.
Vintage Loss Curves
We group your loans by origination month/quarter and track cumulative default rates at matching maturities. If your Q4 2024 vintage is defaulting 50% faster than Q1 2024 at the same point in time, that’s a red flag your model missed.
Population Stability Index (PSI)
PSI measures how much your current applicant population has shifted from the population your model was trained on. A PSI above 0.1 means moderate drift. Above 0.25 means significant drift — your model is scoring a population it wasn’t built for.
Gini Coefficient / KS Statistic / AUC
These measure your model’s ability to rank-order risk — can it still separate borrowers who will repay from those who won’t? A Gini below 0.35 or KS below 25 means your model needs attention.
Roll Rate / Migration Matrix
We track how loans move between delinquency buckets each month. When your Current → 1-30 DPD rate climbs from 5% to 8%, that’s portfolio stress — detectable months before it shows up in default numbers.
Real-Time Alerts. Professional Reports.
When a metric breaches your threshold, we deliver actionable alerts through the channels your team already uses.
CreditVigil
online
⚠️ CreditVigil Alert — [Company Name]
Your scoring model's Gini coefficient dropped from 0.52 to 0.38 over the past 8 months. This means the model is losing its ability to separate good borrowers from bad.
Recommended action: Review score cutoffs and consider model recalibration.
📊 View full dashboard:
https://app.creditvigil.com/dashboard
10:32 AM
Enterprise-Grade Security
We handle sensitive financial data with the care it demands.
Encryption by design
Specified sensitive material is sealed with AES-256-GCM before it is stored. Traffic is intended to be encrypted in transit. Host-level encryption depends on how the service is deployed.
Confidentiality first
A confidentiality agreement may be required before a portfolio file is exchanged. That is an organisational practice, not a software control.
No names required
Default identity handling uses keyed pseudonyms. You do not need to send names or national IDs for monitoring to work. Source files may still contain identifiers until they are classified; analytical tables are designed to exclude direct identifiers.
SOC 2 roadmap
SOC 2 Type II certification is on our roadmap. Internal control checks are mapped to the trust service criteria for security, availability, and confidentiality. Mapping is not a certification.
Ready to see what your models are hiding?
Get a complimentary portfolio health check. We analyze your loan data and show you exactly where your scoring model stands.