Credit Engine
Feed it a raw bureau report. Get back a risk grade, a priced offer, and the exact reasons behind them — deterministic, explainable, and the same every time for the same inputs.
Deterministic, low-latency decisioning. No fabricated benchmarks — we publish numbers once they're measured in your deployment.
Two stages: knock out, then price.
Eligibility is a hard gate. Only files that pass reach pricing, where a transparent scorecard places each applicant into one of ten grades.
Eligibility knockouts
Deterministic pass/fail policy gates — minimum score, age and income bounds, active defaults, recent enquiries, DPD ceilings, FOIR caps. Any knockout fails fast and returns exactly which rule stopped the file.
10-grade risk-based pricing
Passing files cascade grade 1 → 10; the first grade whose gates pass yields the offer. 10 risk grades, each with pricing bands your credit team configures. Rates are set by the lender, not by Gradexa. Same knockouts, same grading, for single-repayment and EMI loan products.
Every decision shows its work.
No score without a reason. Each response carries the rules that fired and the contribution of each factor — ready for adverse-action notices and audit.
{
"grade": 4,
"fired_rules": [
{ "id": "KO_ACTIVE_DEFAULT", "result": "pass" },
{ "id": "KO_MIN_SCORE", "result": "pass" },
{ "id": "GRADE_4_GATE", "result": "matched" }
],
"score_contributions": [
{ "factor": "bureau_score", "points": +48 },
{ "factor": "tradeline_mix", "points": +12 },
{ "factor": "dpd_history", "points": -22 },
{ "factor": "foir", "points": -14 }
],
"offer": {
"limit": 30000
}
}Fired rules
The ordered list of policy rules evaluated and whether each passed, failed, or applied an adjustment.
Score contributions
Signed points per factor, so you can see precisely why a file landed where it did.
Adverse-action ready
The reason codes map cleanly to declined / counter-offer notices your compliance team can stand behind.
One report shape, any bureau.
You pull the report with your own bureau membership and send it on the request. The engine parses it and normalises it into one internal report shape, so your rules are written once and run against any source. Parsers for CIBIL, CRIF High Mark, Experian, and Equifax are built. The engine makes no bureau calls itself; direct bureau pull is on the roadmap.
- 01CIBILparses XMLBuilt
- 02CRIF High Markparses JSONBuilt
- 03Experianparses fixed-widthBuilt
- 04Equifaxparses JSONBuilt
Direct bureau pull
RoadmapToday you pull the report with your own bureau membership and send it; the engine does not call a bureau. Direct pull from the engine — a soft pull at application, a hard pull only before disbursal, configurable per tenant — is on the roadmap.
Formats parsed
CIBIL XML · CRIF High Mark JSON · Experian fixed-width · Equifax JSON. You name the bureau on the request; every report is normalised to one internal shape, so rules are written once.
Stored for re-runs
The normalised inputs of every decision are stored with the policy version in force, so any decision can be re-run and compared. Storing the raw bureau file alongside them is on the roadmap.
Variable dictionary
The rule editor exposes 48 named decision variables: bureau score, credit vintage, enquiries, live loans, DPD history and recency, current overdue, write-off and suit-filed counts, affordability, plus fraud, device and location signals when you supply them. Every variable is named and visible where you write the rule. No black-box features.
Underneath, the engine also derives a deterministic feature set from each report — enquiry velocity, DPD severity and buckets, utilisation, secured mix, vintage — which feeds the transparent score whose contributions come back on every decision.
Run a real bureau report through it.
Bring a scenario. We'll show you the knockouts, the grade, the priced offer, and the reasons — live.