Kreedite
KreeditePlatformCredit Score Generation
Platform

A composite score, composed - never conjured.

The composite score is a weighted average of consented, verified signals. Each signal contributes a raw sub-score, each sub-score is weighted, and the resulting composite lands on a 0-100 band that maps into your policy tiers.

Reference instrument
  • Score band0 - 100
  • Weighting modelKreedite-01
  • Refresh cadenceOn application
The composition

Weighted average, not a mystery function.

The engine deliberately uses a transparent weighted-average composition on top of validated sub-scoring models. This trades a small amount of theoretical accuracy for something more valuable to a lender: defensibility.

  1. 01
    Sub-score each signal

    Each consented signal is scored on a 0-100 scale by a validated sub-model tuned against realized loan outcomes in comparable segments.

  2. 02
    Apply calibrated weights

    Each signal enters the composite with a weight tuned against your book. Weights are static per model version and disclosed to your credit committee.

  3. 03
    Compose the score

    Composite = sum of (sub-score * weight). The band is 0-100 by design, so the mapping to your policy tiers stays simple.

  4. 04
    Attach the attribution

    The score is returned alongside every contributing signal, its sub-score, its weight and its resulting contribution - not a single opaque number.

TRANSACTION PATTERNWeight 31%COMMUNITY REPAYMENTWeight 29%UTILITY PAYMENTSWeight 22%MOBILE RECHARGEWeight 18%CREDIT SIGNAL SCORE75out of 100Approve for review
Worked example - one borrower
SignalSub-scoreWeightContribution
Transaction Pattern7831%24.2
Community Repayment8429%24.4
Utility Payments7122%15.6
Mobile Recharge6218%11.2
Composite score75
Numbers shown are illustrative. Live scoring uses your calibrated weights and the applicant's consented, verified signals.
Score properties

What the number does and does not claim.

A relative signal, not an absolute verdict

The composite score is calibrated to rank order applicants within a comparable segment. It is not a universal creditworthiness rating.

Stable across missing sources

If a signal is unavailable, weights re-normalize across the remaining consented signals. Missing data does not silently penalize an applicant.

Auditable end to end

Every scored decision retains the exact signal set, weights and sub-scores used - so any past decision can be reproduced or defended.

  • Composite band0 - 100
  • Default tier thresholds68 approve / 45 review / below 45 risk
  • Weight drift on missing signalsRe-normalized across available sources
  • Audit trail retentionFull attribution per scored decision
Next step
Score a real borrower cohort - not a demo slider.

Bring five or ten anonymized profiles. We will run the engine end to end and walk your team through the attribution row by row.

See explainability layerBook a working session