See the portfolio move - before it moves you.
Individual scores answer whether one applicant should be lent to. Portfolio monitoring answers whether the underwriting policy is still doing what it was designed to do. Kreedite tracks how your scored cohort behaves across tiers, segments and time.
- Cohort granularityBy product, tier, segment
- Correlation windows30 / 90 / 365 days
- Drift alertsSegment aware
Four cohort behaviors, watched continuously.
Portfolio monitoring is not a monthly PDF. It is a live view of how the scored book is behaving - so a policy owner can respond to drift before it turns into loss.
Realized default rate is tracked per score band. If the correlation weakens, the weighting model has drifted - and it is worth re-calibrating before writing more of that exposure.
How many borrowers move between approve, review and risk tiers over time. A widening review tier is a signal that the scoring band is losing separation.
Whether the composition of the incoming applicant pool has shifted - by geography, by product, by cohort size - relative to the segment the weights were tuned on.
If a signal source (say, community repayment) starts becoming unavailable across more applicants, the composite is quietly relying on the remaining sources. Monitoring surfaces this before it becomes a blind spot.
A view for the policy owner. A view for the underwriter.
Portfolio monitoring is not one dashboard - it is a small set of role-specific views, each answering a specific question that role will actually ask.
Watch tier migration, band separation and segment drift week over week. Trigger a weights review before a soft trend becomes a hard loss.
See realized default correlation against each score band across product lines - a single view for board reporting and regulatory conversations.
Monitor queue mix, review tier growth and reason-code frequency - so the review workflow scales alongside the book.
Track exposure movement between tiers, so provisioning models can be updated on evidence, not on quarter-end panic.
- Default correlation windows30 / 90 / 365 days
- Segment attributes trackedProduct, geography, income band
- Alerting surfaceEmail, webhook, in-product
- Export formatsCSV, JSON, audit-signed PDF
- Every metric is decomposable back to the individual scored decisions that produced it.
- Drift alerts are calibrated per lender - noise floors respect the volume of your book.
- Model version and rule version travel alongside every metric - reproducibility over time is preserved.
We will demo the exact views most useful for your role - policy, underwriting, risk or finance.