CREDIT RISK
Loan default prediction: see which loans are at risk
Loan default prediction with TIMi estimates, for each loan already granted, the probability that the borrower stops repaying as agreed, and shows the factors behind it.
Default probability
Per loan, at a horizon you choose: 30, 60, 90 days or more.
Early warning signals
Late payments and changes in the borrower’s situation feed the model.
Explained per file
Each probability comes with the factors behind it.
HOW IT WORKS
How loan default prediction works.
Three steps, in one engine, on a standard server with no cluster.
1. Prepare the data
Anatella gathers the repayment history of the loan, the borrower’s updated financial data and early warning signals such as delays or changes of situation, then structures them into variables.
2. Build the model
Modeler compares the profile and repayment behavior of the borrower with those of borrowers who defaulted or not, and estimates a probability of default at a given horizon.
3. Score and explain
Each loan gets a probability of default with the factors behind it, such as recent delays, a drop in income or a rising debt ratio.
GETTING IT RIGHT
Choices that shape a default model.
The model is only as useful as the question it is asked.
Choose the horizon
The horizon follows the business goal: 30, 60, 90 days or more, adjusted to how the portfolio is monitored. The target says whether the loan defaulted within that horizon after the observation date.
Look for weak signals
The aim is to spot changes in behavior or in financial situation that may come before a first declared incident.
Separate granting from follow-up
Credit scoring evaluates an applicant before the loan. Default prediction follows a loan already running, with its actual evolution since disbursement.
EXPLAINABILITY
A probability says how likely. The factors say what to do.
A default probability alone gives a level of risk, not what drives it. A white-box model shows the factors for this file, which allows a suitable action (support, restructuring, provision) instead of a uniform response.
For risk and collections teams
High-risk files can be prioritized with a precise context on the cause of the risk.
For transparency requirements
A model that justifies each result variable by variable gives you the documentation to answer transparency expectations such as GDPR and the AI Act. The compliance assessment itself stays with your legal team.
For model owners
A borrower’s situation evolves, so the score is recomputed regularly. Modeler’s Drift Report shows when the model itself needs an update.
DATA
What data a default model uses.
Anatella connects to more than 80 sources, so this data can be gathered without prior custom development.
Repayment history
Payments, delays and incidents.
Updated financial data
The borrower’s current situation.
Early warning signals
Change of professional situation, other credit incidents.
Observed defaults
Past defaults, as a learning base.
PROOF
Same models, applied to risk.
We do not yet have a published customer story dedicated to loan default prediction. These deployments apply the same models to risk in banking and insurance.
98% accuracy on business bankruptcy within 6 months
Built from public balance sheets, with 24,000 variables including many year-over-year evolutions.
€250 million more could be invested safely
An individual-level model of cash-out probability showed that AXA could invest an additional €250 million without increasing liquidity risk.
FAQ
Common questions.
What horizon does the model predict?
The horizon depends on the business goal (30, 60, 90 days or more) and can be adjusted to the way the portfolio is monitored.
How does it differ from credit scoring?
Credit scoring evaluates an applicant before the loan is granted. Default prediction follows a loan already running, taking into account its actual evolution since disbursement.
Can it detect a default before the first late payment?
That is the aim: identifying weak signals, such as a change in behavior or a financial deterioration, that may come before a first declared incident.
Does the score need to be recomputed regularly?
Yes, a borrower’s situation evolves. Modeler includes a Drift Report that shows when a model has degraded, so you update it when needed instead of on an arbitrary schedule.
RELATED USE CASES
Keep exploring.
SEE FOR YOURSELF
Test it on your own data.
Run TIMi on your own data with a free trial, or talk to us about your case.
