Product 07 · Borrower Profile Prediction

A borrower profile with reasons, not just a score.

Before disbursement, Borrower Profile Prediction combines the application, the confirmed income report and how similar borrowers have actually repaid, to describe the behaviour to expect, and explains every factor it used.

Borrower profile · Perera, K. A. N. · personal loan LKR 1.2 M / 36 mPre-disbursement

EXPECTED PATTERN

Reliable, payday-aligned

Like 1,840 similar borrowers, 94% of whom stayed current

AFFORDABILITY BAND

Comfortable

DTI 22% → 41% with this instalment

SUGGESTED TERMS

Due on the 5th

Salary lands on the 25th–28th

FACTORS · WHAT MOVED THE PROFILE

Salary regularity 12/12
+ strong
Similar-profile history
+
Existing leasing obligation
Two unexplained inflows
− minor
Approve as proposedAdjust termsDecline with reason

What it predicts

Four things a credit officer wants to know before saying yes.

Expected repayment pattern

Reliable, payday-aligned, seasonal or irregular: the same vocabulary the Settlement Pattern Analyst uses on the existing book.

Affordability band

Debt-to-income before and after the proposed instalment, from the confirmed income report.

Suggested date and tenor

An instalment date that follows the salary, and a tenor that keeps the band comfortable.

Early-warning sensitivity

Which changes to watch for on this account once it is live, so monitoring starts on day one.

Explanation first

A profile you can defend to the customer and the committee.

1

Gather inputs

The application, the confirmed KYC record and the confirmed income report. Nothing unconfirmed enters.

2

Compare

Against how borrowers with similar income shape, obligations and tenor repaid in your own book.

3

Explain

Every factor is listed with its direction and weight, in plain words the officer can read to the customer.

4

Officer decides

Approve, adjust terms or decline, with a reason recorded alongside the profile.

5

Monitor

The account’s real repayments are compared with the prediction. Misses are reported, and the model is reviewed.

Model governance

Built for your model risk committee.

A prediction that cannot be examined should not be used to lend. The profile ships with the documentation, monitoring and controls a regulator would expect.

No protected attributes

Ethnicity, religion, gender and similar attributes are excluded from the model and its inputs. Proxies are tested for.

Every factor is challengeable

Officers can dispute a factor and record why; disputes feed the review.

Trained on your book only

Comparisons use your institution’s own repayment history. Nothing is pooled across customers.

Drift monitoring and periodic review

Prediction vs. outcome is reported monthly; the model is re-validated on a schedule you set.

Documentation pack

Inputs, method, validation results and limitations, written for your committee and your regulator.

Who decides

The profile informs. The officer approves.

No automatic approvals or declines. The profile is an input to a human decision.
Decisions carry reasons. Approve, adjust or decline, each recorded with the officer’s note.
Customers can be told why. The factors are written to be read aloud.

Works with

Test it against decisions you already made.

Give us last year’s approved loans and how they repaid. We’ll show the profiles we would have written on day one.