Income that lands when it is supposed to
The interval between deposits, not just the amount that arrives.
KoraScore
Predict repayment, not just risk
KoraScore is fit on funded loans that performed or did not, so it predicts real repayment behavior.
One file · 12 months
Trusted by auto and personal lenders holding their loss line while their book grows.
Four patterns the account shows before a credit file does.
The score is built from the same behavior the risk flags read, so the two agree by construction rather than by luck.
A balance that holds through the month
How far the account runs down before the next deposit catches it.
How often the account runs out of money
Counted month by month, so a bad quarter reads differently to a bad year.
Borrowing that never reached a credit file
Payday, cash advance, BNPL and earned wage access, read off the feed.
Charge-off and the first missed payment are two different questions.
Both come back on the same run, so you underwrite on the one your book is measured by instead of conflating them.
Run the model your book is measured on
Both land on the familiar 300 to 850 scale, from one analysis.
See which behavior each model leans on
Each weight is earned on loans that went on to perform.
Read what moved this file, not just its number
The factors come back with the score, in plain language.
Prove it on outcomes you already know.
Backtesting is free, and it runs against your own historical portfolio.
See the lift on loans you have already closed
Your own outcomes, scored as if the file were new.
Or a model built on your portfolio, end to end
Development, backtesting and validation included, then deployed and drift-monitored.
Want the whole file the score is computed from?
Cash Flow AnalysisNeed the behavior flags that sit beside it?
Improve portfolio performanceTrying to approve more, not fewer?
Approve more good borrowersQuestions a model risk committee asks first.
What has to hold before a cash-flow score is allowed near a credit decision.
01What do the two KoraScore models predict, and what do they not?
One’s trained to predict charge-off, the other to predict hitting 60+ days past due. Both score the applicant in front of you on the same 300-to-850 scale you’re already used to. But neither one hands you a decision, a recommendation, or tells you where to queue the file. What that score actually means for a deal is up to your policy, not ours.
02How is a KoraScore explained to a reviewer?
Every score comes back with the factors that moved it, in plain language, plus the underlying attributes behind those factors. So your team can actually check the reasoning instead of just trusting the number.
03Does KoraScore replace the credit score we already run?
Not really, it comes in as its own endpoint alongside whatever you’re already running, and there’s a version built for applicants with no bureau history. Where you slot it into your policy is your call, and your own backtest is what’ll show you the right answer.
04How often is the KoraScore model reviewed once it is live?
We check score distribution, approval-rate trends, and vintage curves every quarter, and keep an eye on population and feature stability in between those check-ins. Thresholds get recalibrated as your book changes.
05Can a file be scored again with KoraScore?
Yes. Score it again whenever you’ve got something new, and we’ll keep every run as its own record rather than replacing the last one. That way you’ve always got a clear trail of what the score was at each point in the file’s life.
Run the score on loans you already know the outcome of.
Send a vintage that has finished performing. We score it as it looked on the day, and set both models next to what actually happened.
We work with auto and personal lenders across the US & Canada