The credit decisioning API built for thin-file
Four signal streams derived entirely from bank transaction data. No bureau pull. One REST endpoint returning a score, a decision, and the factors that drove it — in under a second.
Signal Architecture
Four inputs. One score.
Every model input is derived from bank-permissioned transaction data. No traditional credit bureau data required at any stage.
Cash-flow velocity
24-month transaction inflowsThe rate of change in monthly inflows and outflows over the past 24 months. Rising velocity with low variance is the strongest single predictor of repayment capacity we measure. A borrower whose income has grown steadily over two years is fundamentally different from one with flat or declining inflows — even if bureau models score them identically.
Predicts: Ability to service new credit obligations, income trajectory risk, repayment capacity at 12 and 24 months.
Recurring payment consistency
Monthly obligation cadenceCadence and regularity of recurring payment obligations — rent, utilities, subscriptions, insurance. Consistent on-time timing signals financial discipline with direct credit relevance. A borrower who has paid rent on the 1st of every month for 24 months is demonstrating a creditworthiness signal bureau models systematically ignore.
Predicts: Loan payment punctuality, default probability at 30, 60, and 90 days.
Income irregularity scoring
Inflow variance + seasonalityMeasures inflow variance, seasonality patterns, and gap frequency. Gig economy and self-employed borrowers have inherently irregular income that bureau models penalize. Our irregularity score distinguishes structural income volatility (higher risk) from predictable seasonality (manageable) from temporary gaps (recoverable).
Predicts: Stress tolerance under income disruption, appropriate loan sizing for non-salaried borrowers.
Transaction graph centrality
Counterparty network structureGraph-structural features of the borrower's transaction network — counterparty diversity, payment centrality, merchant category mix, employer transaction stability. Economic integration signals: a borrower transacting with many stable counterparties shows a degree of economic embeddedness that predicts lower default risk independent of income level.
Predicts: Economic stability indicators, long-term default risk, loan size appropriateness.
Decision Output
Every response field, documented
The decision payload is designed to plug directly into your adverse action workflow. Every factor that drove the decision is named and ranked.
The Lendiro cash-flow score. Calibrated to a familiar 300–850 scale for easy integration with existing underwriting thresholds. Higher = stronger repayment signal.
Hard decision output based on your configured score thresholds. REVIEW triggers a manual queue. Thresholds are configurable per integration.
Ranked list of the features that most influenced the decision — positive and negative. Required for ECOA-compliant adverse action notice generation on DECLINED outcomes.
Model confidence given the available data depth. Low confidence (thin data window) may indicate the REVIEW path is appropriate regardless of score.
Pre-formatted adverse action reason codes. Pass directly to your notice generation system. Compliant with ECOA Regulation B notice requirements.
{
"lendiro_score": 742,
"decision": "APPROVED",
"confidence_band": "high",
"lookback_months": 24,
"primary_factors": [
"recurring_payment_consistency",
"inflow_velocity_trend",
"low_overdraft_frequency",
"counterparty_diversity"
],
"adverse_action_codes": [],
"latency_ms": 847,
"request_id": "req_9c4f1a2b"
}Integration Path
Three steps from key to first decision
Drops into your existing underwriting stack via REST. No model retraining. No batch processing cycle.
Connect bank data
Connect via open banking API (Plaid, MX, Finicity) or pass a bank data token your platform already holds. Lendiro ingests the last 24 months of transaction records.
POST to /v1/decision
Send a JSON payload with applicant ID, bank data token, lookback window, and requested loan amount. The REST endpoint accepts standard JSON — no SDK required.
Parse and route
Decision returns in under one second. Read decision to route to approve/review/decline flow. Use primary_factors directly for adverse action notice generation.
Data Handling
What Lendiro does not store
- Raw transaction records are not retained after feature extraction. The Lendiro model operates on derived features, not stored PII.
- Applicant name, account number, and routing number are never written to Lendiro storage. Bank data token is ephemeral.
- Decision records (score, decision, factors) are stored for 90 days for dispute resolution and regulatory audit purposes, then purged.
- No data is sold to or shared with third parties. We process data solely to return the decision response to your platform.
- Lendiro is not a consumer reporting agency. The Lendiro score is a lender-side risk tool, not a consumer credit report. FCRA obligations attach to your platform's use of the decision, not to Lendiro's computation of it.
Fair Lending Posture
Designed with fair lending principles
- Model features are limited to financial behavior signals: cash-flow velocity, payment cadence, income patterns, transaction graph structure.
- Explicitly excluded features: zip code, census tract, merchant geolocation, surname, or any feature that could serve as a demographic proxy under ECOA Regulation B.
- Adverse action reason codes are returned with every DECLINED decision, enabling ECOA-compliant adverse action notice generation without additional effort.
- Model fairness testing is conducted across demographic segments on validation data. Results are available to qualified integration partners under NDA.
Ready to see a live decision?
Get an API key for the sandbox environment in minutes. Full production access follows a brief integration review.