The manager who ran a Rs 12,000 crore large and mid cap fund for 11 years, adding 3.5% a year over the benchmark, resigns; the successor has a four-year record at a smaller fund with 1.2% a year. A client holds Rs 45 lakh in the fund with Rs 18 lakh of unrealised gains. Hold, watch or exit?

Watch, with the new money redirected and a written tripwire; do not exit today. The record belonged to the person who left, so the fund's edge is now a question. But exiting costs about Rs 2.1 lakh of tax for certain, 4.7% of the holding, while the cost of staying is a guess: about Rs 1.0 lakh a year if the successor's 1.2 points replace 3.5. Stop the SIP into the fund, send it elsewhere, and set a twelve-month review with named triggers.

When a restaurant's chef leaves, the room, the menu and the suppliers stay, but the reason you booked may have walked out of the door. A fund's record is evidence about the people and process that produced it, and the person who produced ten winning years out of eleven is gone, so the record is now history rather than a forecast. What remains is the process, the analysts and the AMC's culture, and the successor's own record: four years at a smaller fund in another category, 1.2 points a year. That is a real record, and a modest one, and it is the right thing to judge the fund on now.

Put the two side by side, because they are different kinds of number. Leaving: Rs 18 lakh of gain less the Rs 1.25 lakh exemption, at the assumed 12.5%, is about Rs 2.1 lakh, 4.7% of the holding, paid on the day. Staying: if the successor delivers 1.2 points instead of 3.5, the client gives up about Rs 1.0 lakh a year, so the tax equals roughly 2 years of that gap, and the gap itself is a guess built on four years of someone else's data. The successor still has a positive record; the fund has not become bad, it has become unknown. A certain 4.7% against an uncertain 2.3% a year is not a trade to make in a hurry.

Not waiting and hoping. Three things happen today. The SIP into this fund stops and goes to another fund in the category, which costs nothing and stops the client adding to an unknown. A twelve-month review is written down with triggers that would end it early: a change in the fund's style or turnover, outflows of more than about 15% of assets as other investors leave, key analysts following the manager out, or the fund trailing its benchmark by more than about 3 points over the year. And the AMC is asked the questions that matter: who else is leaving, what changes in the process, and whether the successor keeps the same analyst team. If the fund lags by 2 points a year while she watches, that costs about Rs 0.9 lakh a year, so even 2.3 years of watching costs less than the tax; watching is cheap.

If the fund's edge was visibly the one person: high turnover, concentrated bets, and a successor whose previous fund looked nothing like this one. Or if the client would be exiting anyway for her own reasons, such as rebalancing, in which case the tax is coming regardless and this is the moment to pay it. Even then, sell in slices across tax years so that each year's exemption is used, rather than in one order that pays tax on the lot. Say the limit too: the successor's four years at a Rs 1,500 crore fund tell you little about running Rs 12,000 crore, in either direction, and the only cure for that is time.