AQS turns every address into a risk intelligence node — application velocity, identity patterns, and applicant history in one API call. Your LOS sees a form field. AQS sees the full picture.
{
"address_quality": {
"total_score": 72, "risk_tier": "MEDIUM_CONFIDENCE",
"component_scores": { "pincode": 20, "locality": 15,
"specificity": 12, "nominatim": 15, "gibberish": 10 }
},
"identity_signals": {
"match_type": "pan_hash_exact",
"match_confidence": 0.95
},
"address_risk_graph": {
"velocity_score": 0.6,
"velocity_trend": { "acceleration_flag": true }
},
"applicant_history": {
"ghost_risk_score": 0.25,
"repeat_applicant": true
},
"combined_signals": {
"combined_risk_score": 0.35,
"signal_severity": "medium"
},
"locus": { "household_confidence": 0.85 }
}
They trigger your bureau pull, consume KYC resources, then vanish. Sunk scrub cost with zero recovery.
Husband applies on Flipkart, wife applies on Amazon. Both are legitimate — but the NBFC doesn't know the household now has 2× the EMI burden. Each application arrives as an isolated event.
Multiple PANs at the same geocoded address in 90 days. Joint family or synthetic cluster? Without address intelligence, you can't tell.
"11 applications in 90 days" looks the same whether it's steady or a sudden spike. The acceleration is the real signal.
| Signal | What AQS detects | Severity |
|---|---|---|
| Velocity acceleration | 1 app/month → 11. Acceleration ratio 2.93×. Dormant-to-active. | High |
| Identity concentration | 6 different PANs at same address in 90 days. Score discounted for multi-unit. | High |
| Address quality jump | Same PAN seen before at score 35. Now back at 82. | Elevated |
| PII rotation | Husband PAN + wife phone → abandoned. Now wife PAN + husband phone. | Medium |
| Dealer geo-mismatch | Dealer in Pune, 4 of 10 apps 300km+ away. Pincode-to-pincode distance. | High |
{
"address_quality": {
"total_score": 72,
"risk_tier": "MEDIUM_CONFIDENCE",
"component_scores": {
"pincode_city_state": 20, "locality": 15,
"specificity": 12, "nominatim": 15, "gibberish": 10
},
"flags": ["LOCALITY_FUZZY_MATCH"]
},
"identity_signals": {
"match_type": "pan_hash_exact", "match_confidence": 0.95,
"linked_keys": { "pan": 1, "phone": 2, "email": 1 },
"address_loci_linked": 3
},
"address_risk_graph": {
"facts": { "applications_90d": 5, "unique_profiles_90d": 2 },
"scores": { "velocity": 0.6, "identity_rotation": 0.3 },
"velocity_trend": { "trend_direction": "accelerating", "acceleration_flag": true }
},
"applicant_history": {
"prior_checks": 2, "ghost_risk_score": 0.25,
"repeat_applicant": true, "days_since_last_check": 15
},
"combined_signals": {
"combined_risk_score": 0.35, "signal_severity": "medium",
"process_signals": ["repeat_applicant"],
"address_signals": ["high_identity_count_at_address"]
},
"locus": {
"household_confidence": 0.85,
"multi_unit_flag": false, "commercial_flag": false
}
}
Free-text in your existing form, or via the optional AQS Address Capture Widget for structured, validated input.
→POST /api/v1/score with address + HMAC-hashed PAN/phone. Raw PII never leaves your systems. One call.
→Address quality, identity signals, risk graph, applicant history, combined signals, locus — all in <200ms.
→Proceed to bureau? Step-up verification? Flag for review? Your rules, your playbook. AQS provides the evidence.
HMAC-keyed hashes only. The NBFC controls the key. We are architecturally unable to identify your customers.
Keyed-hashed identifiers with lender-managed keys. Geocoordinates from pincode centroids. Compliance review in minutes.
Facts, scores, reason codes — never recommended actions. Your policy engine maps signals to decisions.
Every response carries rules_version, scoring_version, and feature flags. Governance metadata answers "why did this score change?"
Each lender's data strictly isolated. No cross-lender signals unless consortium agreement.
No identity block? Same response as today. Premium layers activate only when hashed PII is provided.
30-day pilot. Observation mode first. We show you what your origination pipeline is missing — then you decide.