Layer 4 of the decision stack. Pull the levers the business actually debates — rate, retention, leakage recovery, expenses, reinsurance cession, investment yield, bancassurance growth — and see the combined ratio, PAT, underwriting result and solvency recompute against the audited FY25 book.
Every figure here is arithmetic, not generated text. Move a slider and the model recomputes synchronously from the reconciled baseline — the same levers always produce the same answer.
Price on renewal. Elasticity gives back 35% of the gain as lapse.
Change in policy retention, in percentage points of GWP.
Fraud, waste and abuse recovered as a share of incurred claims.
Change in management expenses. Negative is cost out.
Additional cession of NEP. Sheds claims and required capital with it.
Change in yield on the ₹7,938 Cr book.
New GWP from the Central Bank channel.
Gross insured CAT loss. 70% recovered from the XoL treaty; 10% reinstatement on the recovery.
One combined-ratio point is ₹37.5 Cr on net earned premium. Baseline is the 112.58% these components actually foot to — +0.58 pts against the published 112% headline, which is rounding in the book, so a neutral scenario shows zero change. Claims ₹2,961 Cr · commission ₹550 Cr · expenses ₹714 Cr · investment income ₹602 Cr.
Stated openly, because a simulation whose assumptions are hidden is just a number generator.
Why this replaced the old panel: the earlier decision stack asked a language model for simulation numbers. Those figures had no mechanism behind them and could not be reproduced. This engine is auditable line by line — and it flags its own limits, warning when a lever is pushed past what any published benchmark supports.