Inputs behind better credit decisions.
Saarathi interprets alternative financial signals from verified official sources to show how a small business actually manages its money.
Four alternative financial data categories.
Each source answers a different question about a micro-business's financial management. Together, they show details that bureau scores miss.
Bank cash flow
A 90-day account view shows income stability, payment habits, overdraft frequency, and balance behavior. It is the strongest single source for near-term repayment capacity.
GST filing pattern
Regular GST filings indicate business formality, steady revenue, and disciplined tax compliance. Missing or inconsistent filings can reveal financial pressure before other signals do.
UPI activity density
UPI volume, counterparty range, and merchant mix show the diversity and stability of business revenue and supplier relationships.
Invoice data
Receivable age, buyer concentration, and invoice frequency provide distinct credit signals. A business with varied buyers and short collection cycles carries a different risk profile from one reliant on a single major client.
Consent-led collection, API-connected to your workflow.
Borrower consent is recorded before processing begins. Data moves through verified official channels using the Account Aggregator framework.
Technical specifications
- REST API, JSON responses
- Average latency below three minutes, based on benchmark runs
- Account Aggregator support
- Webhook for real-time decisions
- API documentation and sandbox access included
Consent architecture
Each retrieval begins with an NBFC-side consent record tied to the borrower's application. Saarathi neither stores nor reuses raw source data after assessment completion. This design follows the purpose-limitation requirements of the DPDP Act 2023.
Data checks before scoring.
Reliable models require reliable data. Every source is verified during collection before entering the feature pipeline.
AA-verified bank data
Bank data comes through the Account Aggregator framework, directly from the bank's systems rather than a borrower-submitted PDF. Self-submitted PDFs are excluded from scoring.
GSTN direct retrieval
GST data is retrieved directly from the GSTN portal and timestamped. Manually uploaded summaries are rejected; data must come through the official portal connection.
Cross-signal consistency check
Before reaching the model, all signals are checked against one another. For instance, high bank-reported revenue without matching GST revenue creates a consistency flag, not a score.