Use case #0002

Margin simulation: what Risk-Based Pricing AI models before every offer

A rate offer is not a pricing decision — it is an NIM commitment. An institution that offers 9.60% to acquire a borrower without modelling the expected credit cost, the funding cost at current rates, the processing cost of this loan type, and the expected prepayment behaviour is not pricing — it is guessing. The Risk-Based Pricing Agent AI runs a margin simulation for every rate offer before it is made, confirming that the rate produces an acceptable net interest margin even at the expected loss rate for this specific borrower profile.

A rate offer is not a pricing decision — it is an NIM commitment. An institution that offers 9.60% to acquire a borrower without modelling the expected credit cost, the funding cost at current rates, the processing cost of this loan type, and the expected prepayment behaviour is not pricing — it is guessing. The Risk-Based Pricing Agent AI runs a margin simulation for every rate offer before it is made, confirming that the rate produces an acceptable net interest margin even at the expected loss rate for this specific borrower profile.

The margin components that the pricing model must balance

The net interest margin (NIM) on a retail loan is not the difference between the rate and the cost of funds. It is that difference, reduced by the expected credit cost (the probability of default multiplied by the loss given default), further reduced by the processing and operational cost of the loan, and adjusted for the prepayment risk that shortens the life of high-rate loans. A pricing model that only looks at the rate-to-cost-of-funds spread will consistently underprice credit risk and overprice competitive risk — both errors cost the institution.

The four margin components that the Pricing AI models for each offer are: funding spread (the rate offered minus the institution's cost of funds at the current tenor), expected credit cost (probability of default for this borrower's risk tier multiplied by the expected loss given default), operational cost allocation (processing, collection, documentation, and compliance cost amortised over the expected loan life), and prepayment risk adjustment (higher-rate borrowers who prepay when rates fall reduce the institution's realised margin below what the original rate implied). All four are computed before the offer is generated.

"An offer that wins the borrower but loses on NIM is not a good offer — it is a subsidised acquisition. The Pricing AI does not make offers that win customers and lose money."

The margin simulation: three rate scenarios for a single borrower

Margin Simulation — Application LA-2025-8841 · Priya Ramachandran · ₹28L Home Loan · 20yr
Three rate scenarios modelled · Recommended rate highlighted · Nov 14, 2025
Scenario A — Aggressive (9.50%)
Offered rate9.50%
Cost of funds7.40%
Gross spread2.10%
Expected credit cost0.68%
Operational cost0.42%
Prepayment adj.0.18%
Estimated NIM0.82%
NIM floor1.00% (minimum)
ResultBelow floor · Rejected
Scenario B — Recommended (9.75%) ✓
Offered rate9.75%
Cost of funds7.40%
Gross spread2.35%
Expected credit cost0.68%
Operational cost0.42%
Prepayment adj.0.16%
Estimated NIM1.09%
NIM floor1.00% (minimum)
ResultAbove floor · Approved ✓
Scenario C — Conservative (10.10%)
Offered rate10.10%
Cost of funds7.40%
Gross spread2.70%
Expected credit cost0.68%
Operational cost0.42%
Prepayment adj.0.22%
Estimated NIM1.38%
NIM floor1.00% (minimum)
ResultAbove floor but above competitor rate · Conversion risk
Scenario A (9.50%) is below the institution's 1.00% NIM floor — the credit cost and operational cost cannot be covered at this rate. It is not offered regardless of competitive pressure. Scenario C (10.10%) exceeds the competitor rate for this CIBIL band by 25–50bps — the conversion probability drops significantly and the incremental NIM does not compensate for the volume lost. Scenario B (9.75%) produces an NIM of 1.09% — 9bps above the floor, competitive with the market, and appropriate to Priya's specific risk profile. This is the offer generated.

The four margin components: how each is computed

The expected credit cost is computed from the borrower's risk tier and the institution's historical loss rate for that tier. For a CIBIL 720–739 salaried home loan borrower, the institution's portfolio data shows a 0.72% annualised credit cost (0.68% used above, adjusted for this borrower's specific profile including their income trend). This is not a generic industry figure — it is the institution's own loss history for this specific risk cell, updated quarterly.

The operational cost allocation is the total cost of originating, servicing, and collecting a loan of this type and tenure, divided by the expected outstanding balance over the loan life. For a ₹28 lakh home loan at 9.75% over 20 years, the origination cost (legal, processing, KYC, technology), the annual servicing cost (statement generation, NACH management, annual review), and the expected collection cost (based on portfolio delinquency rate) total approximately ₹1.18 lakhs over the loan life — amortised to approximately 0.42% of the outstanding balance annually.

Prepayment risk is the risk that the borrower refinances to a lower rate (or takes a better offer from a competitor) before the expected loan tenure ends. Higher-rate borrowers are more likely to prepay when rates fall, which cuts the institution's realised margin on those loans. The prepayment adjustment accounts for this expected income reduction, which is higher for Scenario C (10.10%) and lower for Scenario B (9.75%) because a competitive rate is less likely to trigger a refinance.

4Margin components modelled — funding spread, credit cost, operational cost, prepayment risk
1.09%Estimated NIM at 9.75% — 9bps above the institution's 1.00% NIM floor for this product
3Rate scenarios tested — Aggressive rejected (below NIM floor), Recommended selected, Conservative noted as conversion risk
NeverOffers generated below the NIM floor — regardless of competitive pressure or acquisition target pressure

The NIM floor is not a constraint on pricing — it is the definition of viable pricing

Every institution has a cost of funds, a cost of credit, and a cost of operations. The sum of those three costs defines the minimum rate at which a loan is economically viable. An offer below that minimum — made to match a competitor, to hit an acquisition target, or to avoid losing a borrower — is not a competitive pricing decision. It is a cross-subsidisation decision: this loan is being partially funded by the margin on other loans in the portfolio. The Risk-Based Pricing Agent AI models all three costs for every offer before it is made, and does not generate an offer that falls below the NIM floor — even when the resulting rate is above the best available competitor rate. The institution that prices correctly sustains its margins across credit cycles. The institution that prices to win at any rate does not.

← Back to Risk-Based Pricing Agent AI