Use case #0001

How Pricing AI adjusts rates dynamically as risk signals change

A single interest rate applied to all borrowers is not a risk management strategy — it is a subsidy from low-risk borrowers to high-risk ones. The borrower with a CIBIL score of 790, 36 months of on-time payments, a stable salaried income, and an LTV of 62% is paying the same rate as a borrower with a CIBIL of 682, two DPD events in the last 18 months, and an LTV of 74%. The Risk-Based Pricing Agent AI ends that subsidy — and does so without manual rate-card management.

A single interest rate applied to all borrowers is not a risk management strategy — it is a subsidy from low-risk borrowers to high-risk ones. The borrower with a CIBIL score of 790, 36 months of on-time payments, a stable salaried income, and an LTV of 62% is paying the same rate as a borrower with a CIBIL of 682, two DPD events in the last 18 months, and an LTV of 74%. The Risk-Based Pricing Agent AI ends that subsidy — and does so without manual rate-card management.

What dynamic pricing actually requires — and why static rate cards fail

A static rate card assigns a rate to a credit tier: CIBIL 750+ gets 9.60%, CIBIL 700–749 gets 10.10%, CIBIL below 700 gets 10.75%. This is simple to administer and easy to explain — and it systematically misprices both ends of each band. A borrower at 749 is priced identically to a borrower at 702, despite a material difference in their risk profile. A borrower at 751 gets the same rate as a borrower at 789, despite paying a premium in credit quality that the institution is not reflecting in its pricing.

Dynamic risk-based pricing replaces the band with a continuous function: each risk signal contributes a precise adjustment to a base rate, and the sum of all adjustments is the offered rate for that specific borrower. The adjustments are computed in real time from the credit file data, not from a manual lookup table that is reviewed quarterly. When a risk signal changes — a new bureau enquiry, a cash flow score that has improved over the last 3 months — the rate adjusts accordingly.

"A rate card with five tiers prices 20% of borrowers accurately. A dynamic pricing model that adjusts per signal prices every borrower accurately — including the ones at the tier boundaries who were worst served by the old card."

The risk-based rate ladder: how the Pricing AI builds a rate for each borrower

Base rate: 9.50% (RLLR + institutional spread) · Signal adjustments applied above
Tier A
Excellent

CIBIL 760+, 0 DPD (36 months), FOIR <35%, cash flow score 85+, LTV <65%

Exemplary credit profile — lower risk than institution average. Rate reflects the quality premium the borrower has earned through consistent performance.

9.50%
Base rate · No adjustment · Best rate available
Tier B
Strong

CIBIL 720–759, 0 DPD (24 months), FOIR 35–42%, cash flow score 75–84

Strong credit profile — minor adjustments for one or two signals that are below the Excellent band. Most approved salaried home loan borrowers sit here.

9.85%
Base + 35bps · Score band (+20) · FOIR (+15)
Tier C
Adequate

CIBIL 680–719, 1×30 DPD (outside 12 months), FOIR 42–48%, moderate cash flow

Adequate credit profile — within policy but with visible risk factors. Rate reflects the incremental risk above the Tier B profile without over-penalising borrowers who meet all policy gates.

10.40%
Base + 90bps · Score band (+40) · FOIR (+30) · DPD history (+20)
Tier D
Elevated

CIBIL 650–679, 1×30 DPD (within 12 months), FOIR 45–52%, or thin-file via alternate pathway

Elevated risk — at the margin of policy. Rate compensates for higher expected loss rate. Borrower is viable but priced to reflect the actual credit risk being taken.

11.25%
Base + 175bps · Score band (+80) · FOIR (+45) · Recent DPD (+50)
Refer /
Decline

CIBIL <650, or score 650–679 with 60+ DPD in 12 months, or FOIR >52%

Outside priceable range — the risk premium required to make lending economic at this profile exceeds what the borrower's FOIR can support in EMI terms, or exceeds what can be commercially offered in the market.

N/A
Decline or refer to credit committee

The signal adjustment table: what moves the rate and by how much

CIBIL score band
Primary risk signal — largest single adjustment

Every 10-point band below 760 adds to the rate

760+: 0bps · 740–759: +10bps · 720–739: +20bps · 700–719: +30bps · 680–699: +40bps · 660–679: +60bps · 650–659: +80bps. The adjustment is per 10-point band, not per tier — a 741 and a 758 receive different adjustments, unlike in a static rate card.

0–80 bps
DPD history
Payment behaviour — recency weighted

DPD events add to rate; recency and severity both matter

0 DPD (36 months): 0bps · 1×30 DPD (13–36 months): +20bps · 1×30 DPD (within 12 months): +50bps · 2×30 DPD: +80bps · Any 60+ DPD: +150bps or refer. A DPD event from 28 months ago receives less penalty than one from 8 months ago — the adjustment reflects recency, not just occurrence.

0–150 bps
FOIR at proposed
Debt service capacity — residual buffer signal

Higher FOIR means less capacity cushion — priced accordingly

FOIR <35%: 0bps · 35–40%: +15bps · 40–45%: +30bps · 45–50%: +45bps. A borrower at 50% FOIR has minimal income buffer above their debt obligations — any income disruption creates immediate default risk. The pricing adjustment reflects this elevated sensitivity to income shocks.

0–45 bps
LTV (secured loans)
Collateral coverage — recovery rate signal

Higher LTV means less collateral cushion in a default scenario

LTV <60%: −10bps (positive credit for over-collateralisation) · LTV 60–70%: 0bps · LTV 70–75%: +15bps · LTV 75–80%: +30bps. A lower LTV provides better recovery if the loan defaults — this is reflected as a rate benefit, not merely a neutral observation. The −10bps is the only downward adjustment in the model.

−10 to +30 bps
Cash flow score
Financial behaviour signal — bank statement derived

A strong cash flow score partially offsets a weaker bureau score

Cash flow score 85+: −10bps · 75–84: 0bps · 65–74: +10bps · <65: +20bps. A borrower whose cash flow score (from the Bank Statement Analyst AI) is materially stronger than their CIBIL score suggests receives a partial rate benefit — the gap between bureau score and cash flow score is an indicator that the bureau score may understate credit quality.

−10 to +20 bps
Income trend
Forward-looking capacity signal — 12-month trajectory

Improving income trajectory reduces the rate; declining increases it

Income growing 15%+ (12 months): −10bps · Stable (±5%): 0bps · Declining 10%+ (12 months): +20bps. A borrower whose income has grown 20% over the last 12 months is a better credit than they were 12 months ago — the pricing adjustment reflects the forward-looking capacity, not just the current snapshot.

−10 to +20 bps

Dynamic rate in action: a real-time adjustment triggered by a new signal

Live Rate Computation — Application LA-2025-8841 · Priya Ramachandran · Home Loan ₹28L
Base rate 9.50% · All signals processed · Rate computed: Nov 14, 2025 · 09:42:11
CIBIL score 736 · Band 720–739 · Adjusts above base rate+20 bps
DPD history: 0 DPD in 24 months (1×30 DPD at 28 months — outside adjustment window)0 bps
FOIR at proposed loan amount: 38.4% · Band 35–40%+15 bps
LTV at requested amount: 68.2% · Band 60–70% · Neutral0 bps
Cash flow score: 84 · Band 75–84 · Neutral0 bps
Income trend: +37% over 12 months (bank statement) · Improving tier — rate benefit applied−10 bps
9.50%
Base rate (RLLR + spread)
+25 bps
Net signal adjustment
9.75%
Offered rate · This borrower
● Offered rate 9.75% · Below competitor average 9.85% for this CIBIL band · Income trend drives −10bps benefit · Rate locked for 90 days
6Risk signals that adjust the rate — each with a precise basis-point contribution, not a tier assignment
9.75%Priya's offered rate — 10bps below the static Tier B rate because her income trend earns a −10bps benefit
+25 bpsNet signal adjustment above base rate — the precise price of Priya's actual risk profile, not her tier's
Real-timeRate recomputed if any signal changes — income data refreshed, new bureau pull, cash flow score update

The rate that wins is the rate that is precisely right — not the rate that is approximately right for a bracket

A Tier B rate of 9.85% applied to every borrower with a CIBIL score between 720 and 759 over-prices the borrower at 758 and under-prices the borrower at 721. The one at 758 compares with competitors and finds they can do 9.75% — and takes their loan elsewhere. The one at 721 gets a rate that does not fully reflect their risk. Both outcomes cost the institution: one in lost volume, one in inadequate risk compensation. The Risk-Based Pricing Agent AI computes 9.75% for the borrower who earns it and 9.90% for the one who does not — from the same base rate, using the same signal adjustments, consistently applied. Every borrower pays the price of their actual risk profile. No more, no less.

← Back to Risk-Based Pricing Agent AI