Built for microfinance realities - not adapted from consumer credit.
Group-lending outcomes, seasonal cash-flow variance, community repayment history and geographic factors are treated as first-class signals - not as afterthoughts inside a scoring model tuned for urban salaried applicants.
- Community signal weightFirst-class
- Seasonality handlingNative
- Field-officer explanationsPlain language
A scoring model that respects how your borrower actually earns and repays.
Most alternative-data scoring tools were built for the salaried, banking-app-using urban applicant, then labelled inclusive. Kreedite was built the other way around - starting from group lending, agricultural cycles and mobile-first economies, then extended outward.
Group-lending and joint-liability history is weighted alongside utility and mobile signals - not tacked on as a hint.
Variance patterns tied to harvest, festival or trading seasons are recognized instead of penalized as inconsistency.
Plain-language reason codes and narratives are formatted for field officers doing manual verification - not for a data-science reader.
Signal availability, weight calibration and drift alerts are configured per district and per branch - because microfinance lending is not one national segment.
- Group-lending signal supportedYes
- Field-officer explanationsPlain language, printable
- Seasonality re-weightingConfigurable per segment
- Branch-level drift monitoringNative
- Language surfacesMulti-lingual narratives (configurable)
Alternative-data credit for underbanked borrowers is not a marketing category. It is the reason this engine exists. The design choices above are what make the difference between a scoring tool that helps a microfinance book and one that quietly optimizes for the borrower it was built around.
In the branch. In the field. In the back office.
A compact view of scored applications with reason codes, contribution bars and manual-override capture - built for a supervisor doing an end-of-day tier review.
A plain-language explanation of the composite is generated for the officer verifying an applicant in the field - printable and shareable within policy.
Segment migration, community-signal availability and default correlation are tracked per branch - so the head office sees which regions the model is starting to lose fit on.
- The recommendation tier is advisory - the lending decision remains with the branch and its policy.
- Consent for each signal source is captured at the branch and retained per applicant.
- Weights and narratives can be localized per language and per region without re-training the underlying model.
A one-branch working session with anonymized cases, walked through together end to end.