Credit decisioning software
that reads the bank account
Balance first credit decisioning on live bank data. Explainable line by line, gated by your program's rules, and tuned on a real small-dollar book. An optional module on the Lendfy platform.
Three steps. Thirty seconds.
Decisioning plugs into Lendfy origination, or into your own flow through the API.
1. Connect the bank data
The applicant links their account through your aggregator (Plaid, MX). Lendfy pulls balances and transactions with consent on record.
2. Score and explain
The balance first engine reads income cadence, balance floors, obligations, and risk events, and writes down why, factor by factor.
3. Your rules decide
The score is a recommendation. Your program's gates, caps, and cooldowns make the call, and the whole trace lands on the audit log.
What the engine actually reads
Credit files are thin exactly where small-dollar lending lives. The bank account isn't. It's the freshest, most honest record of a borrower's month.
Balance trajectory
Floors, swings, and the shape of the month. Does the account breathe, or does it flatline before payday?
Income cadence
Real payroll deposits and their rhythm (weekly, biweekly, semi-monthly), detected from the transactions, not self-reported.
Risk events
NSF and overdraft activity weighted by recency and pattern. One bad week two months ago isn't a life sentence.
Existing obligations
Recurring debits, competing lenders, and payment stacking visible in the flow, before you become the fifth pull on payday.
Timing fit
Where the due date should land relative to the payday. The engine recommends collection timing, not just approval.
Your program's gates
Caps, cooldowns, screening requirements, and state rules sit above the model. Recommendations never bypass them.
Every decision shows its work
A score you can't explain is a liability with a decimal point. Decisioning writes the reasons down for your operators, your adverse-action letters, and your examiner.
Each contribution listed with direction and weight, recorded on the application, forever.
Declines carry structured reasons your letters can cite. No reverse-engineering a black box at complaint time.
The engine is tuned against real repayment outcomes rather than a static scorecard, and every model change is measured before it reaches a decision.
Decisions in seconds, reasons forever
See Decisioning score a live-shaped application end to end: bank data in, explained recommendation out, your rules on top.