Defend a long on a private bank under questioning: NIM 4.1%, ROA 1.6%, gross NPA 2.1%, P/B 3.2x, but unsecured personal loans grew 48% last year to 22% of the book. What three questions will the interviewer grill you on, and what are your answers?

The three questions are: why the unsecured book is growing twice as fast as the bank, what the 3.2x multiple assumes about returns, and what happens to both if unsecured losses rise. Answer with the lag: the 2.1% NPA ratio is diluted by Rs 9,106 crore of young loans. The multiple needs a return on equity of about 16% to hold indefinitely, which is today's 16%. A 5 point rise in unsecured losses cuts ROA to about 0.8% and earnings by about 52%. Hold the long only if you can show the underwriting is different.

A restaurant that doubles its covers by lowering the price of the set menu is busier, not better. A bank that grows unsecured lending at 48% while the book grows 20% is either finding customers nobody else saw, or lending to people other banks turned down, and the interviewer wants you to say which. The honest answer is that the numbers cannot yet tell you, because unsecured loans go bad with a lag of 12 to 18 months. What you can show is the size of the untested book: about Rs 9,106 crore of unsecured loans made in the last two years, 11% of all loans, on which the bank has almost no loss history of its own.

Because growth dilutes it. Suppose 4% of the unsecured loans that are over a year old have gone bad. Those loans are last year's Rs 11,892 crore; this year's additions have barely had a first missed payment. 4% of Rs 11,892 crore spread over today's Rs 17,600 crore reports as about 2.7%, so the ratio understates the loss rate on the loans that have actually been tested. The number to ask for is the vintage loss curveLosses tracked by the year the loans were made, so each year of lending is judged on its own after the same amount of time, instead of being averaged with newer loans.: how each year's loans performed at 12 and 18 months. If the newer vintages are tracking worse, the 2.1% is a photograph of the past.

Turn the multiple into the returns it needs. With assets ten times equity, a 1.6% return on assets is a 16% return on equity, so 3.2x book is about 20x earnings. On a simple growth model with a 12% required return and 10% long-run growth, 3.2x book needs a return on equity of about 16.4% held indefinitely, which is today's 16% with no room for a bad year. That model is crude, and say so: small changes in the growth and return assumptions swing it a lot. But its message is what the interviewer wants to hear from you: the multiple is a bet on ROE, and ROE is a bet on credit costs.

Run the stress out loud. Losses on the unsecured book rise 5 points; on 22% of loans that is 1.1% of all loans in extra credit cost, about 0.83% after tax. Return on assets falls from 1.6% to about 0.78%, return on equity from 16% to about 7.8%, so earnings roughly halve, and even at the same 20 times earnings the stock is worth about 1.6x book, a fall of about 52%. The market rarely keeps paying 20 times for a bank whose growth engine just broke, so that is the kind case. The bank would still be profitable; the stock would not be. That asymmetry is the case against the long at this price, and you should put it on the table before the interviewer does.

Only on evidence the pitch did not yet have. Hold it if the vintage data shows the new unsecured loans are to the bank's own salaried deposit customers, with early delinquencies no worse than older vintages; drop it if the growth came through third-party apps and new-to-bank borrowers. Say what the position is worth either way: at 3.2x book the stock pays nothing for a bad year, so the right size is small until the vintages have aged through one. An interviewer grilling a pitch is not looking for you to win; they are looking for you to know which of your assumptions carries the weight and what would make you change your mind.