Guide · Mutual funds and debt · Updated 2026-09-18

A segregated portfolio does not freeze your money, it freezes the asset nobody can price

A debt scheme holding a bond that has just defaulted has a pricing problem and a queue problem, and the queue problem is the dangerous one. Nobody knows what the instrument is worth, because its worth depends on a recovery process that has not started, yet the scheme must still publish a net asset value every business day and still pay anyone who asks to leave, at that value, in cash, out of assets it can sell. That is a run: not a panic, a piece of arithmetic that rewards speed, and the mechanism is built against it line by line.

The run is the default outcome, and it is arithmetic

Take an illustrative scheme of one hundred crore rupees across ten crore units, a net asset value of ten rupees, with eight crore in one issuer's paper. The issuer is downgraded below investment grade, the valuation agencies mark that holding down by a quarter to six crore, and the published value falls to 9.80 for everybody that evening. So far the system has worked: an honest markdown, borne equally.

The problem is what the 9.80 assumes. Suppose the eventual recovery, years later, is 3.20 crore. A further 2.80 crore of loss is sitting in the scheme unrecognised, and it cannot be recognised because no price exists to recognise it at. Across ten crore units that is 28 paise, so the honest value of every unit is 9.52.

Now let people leave. Redemptions cannot be paid in bonds nobody will buy, so they come out of the liquid assets. Three crore units go out at 9.80, then another three crore, each payment made at a value containing an assumption about the impaired holding and settled with assets that were never impaired. The holding does not shrink: its share of what remains grows from eight percent past fifteen.

When the recovery arrives at 3.20 crore, the 2.80 crore shortfall has nowhere to land except the four crore units still there. That is 70 paise a unit, so they realise 9.10 while the six crore units that left realised 9.80, which is 28 paise each more than their share. Six crore at 28 paise is exactly the 1.68 crore that four crore at 42 paise gave up. Nothing was destroyed by the run. It was transferred, from the slow to the fast.

The framework was written immediately after an episode in which one large financing group's paper travelled from the highest available rating to default inside a few weeks, and debt schemes across many fund houses found simultaneously that they held it. The concern was never a bad bond. It was that one issuer's credit event could force distressed selling of good assets across the whole industry. A run is contagious in a way a default is not.

What the trigger actually is, and who is allowed to pull it

A segregated portfolio cannot be created because a bond is falling, because a sector looks fragile, or because the manager would prefer not to show a mark. The trigger is a defined credit event at the issuer level, and it is objective by design.

Two points sit inside that table. The trigger is at the issuer rather than the instrument, which is why a loan rating downgrade counts even though the scheme holds bonds. And a scheme can end up with more than one segregated portfolio, because a deteriorating credit keeps producing credit events.

The permission is not automatic. The scheme information document must already contain an enabling provision, and the asset management company must have a written policy approved by its trustees. A scheme whose documents are silent cannot invent the power on the morning it becomes useful, which makes this one of the few things about a credit event you can check in advance. Using it remains optional even where it is available.

One record date, two portfolios, the same unitholders

The sequence on the day is tight, and every step closes a window a run would otherwise walk through.

Allotment is mechanical and deliberately unclever. Every investor on the books at the date of the credit event receives the same number of units in the segregated portfolio as in the main portfolio. No proportions change, no discretion is exercised, nobody is chosen.

For reporting, the segregated portfolio is not a separate scheme. Its assets are included in the scheme's, its folios are not counted twice, and the number of segregated portfolios created must appear in the scheme documents, account statements, portfolio disclosures and any advertisement showing scheme performance. That last one matters: the main portfolio's record cannot be shown afterwards as though the segregated portion never existed.

There is a second beneficiary that rarely gets mentioned, the investor who has not arrived yet. Without segregation, somebody subscribing the day after a downgrade buys into an unpriceable hole at a value that assumes it away, diluting the loss for holders already there. After segregation they buy a clean main portfolio at 9.20 and take no part of a claim they never owned.

What the frozen side is worth, and who is allowed to say

The valuation is not the fund manager's opinion. Securities rated below investment grade must be valued at prices provided by the valuation agencies appointed for the industry, so the number on the segregated portfolio is set outside the fund house that created it. A manager marking their own impaired holding is precisely the conflict the framework removes.

The methodology is a graded haircut that deepens as the credit step falls, applied to principal and, at the same percentage, to accrued interest, so a holding that has stopped paying does not keep accruing income at full value into the net asset value. A fund may deviate, but only with a recorded rationale, which turns a judgement call into a documented one.

The table shows the structure, not a current number: actual percentages vary by seniority, security and sector, and they are revised. What matters is that the mark is graded rather than binary, externally supplied, and still only a mark. The segregated portfolio in the worked example carries 0.60 a unit and pays 0.32. That gap is the part nobody could price, and the design exists to land it on the units that owned the exposure.

The fee that cannot be charged, and the framework that renamed it

The frozen side is not a fee-earning asset. No investment and advisory fee may be charged on a segregated portfolio at all, for as long as it exists. Other permitted expenses may be charged only pro rata and only upon recovery, subject to a ceiling referenced to the main portfolio, whose own costs never absorb the segregated side's.

That last row answers the obvious objection. If a fund house could freeze its worst holding, stop showing it in the headline number and carry on collecting a fee, segregation would be a tool for hiding mistakes. It cannot: the fee stops, the disclosure follows the scheme into every performance advertisement, trustees receive action taken reports on the recovery, and the people who bought the instrument lose incentives over it.

The fee rule is where the current position needs care, because the framework it is written against has just been rebuilt. The mutual fund rulebook in force since 1 April 2026 replaced regulations that had governed the industry for three decades, and alongside it the all-in total expense ratio was recast into a base expense ratio with statutory levies and brokerage shown outside it. The substance of the segregation rule survives; its vocabulary and cross-references do not, so every page on side-pocketing written before 2026 cites a repealed rulebook.

Listed, transferable, and still not liquid

Segregated units must be listed on a recognised stock exchange within ten working days of creation, and the asset management company must process transfer requests. That is useful for a transmission or a corporate action, and close to useless as an exit. A buyer is being asked to price a recovery the fund itself cannot price, so any bid carries a discount for that uncertainty on top of a carrying value that already carries a regulatory haircut, in an instrument with no continuing flow of buyers. The listing is a legal route for the units to move, not a market that will price them on demand. A segregated unit is a claim, not a position.

Where a recovery goes, and where it cannot go

Recovery on the impaired instrument is distributed to the holders of segregated units, in proportion to their holdings, as and when it is received. It does not flow into the main portfolio, it does not top up the scheme's published performance, and it is not held back.

The consequence that surprises people is that redeeming the main portfolio does not surrender the claim. An investor who exits the day after the split still holds their segregated units and still receives their share of whatever comes back. Segregation never forces anybody to stay in a fund they have lost confidence in. It separates leaving the fund from a claim they already owned.

A write down to zero is also not the end. A segregated portfolio can be marked to nothing well before any resolution concludes, and anything recovered afterwards is still distributed, sometimes in instalments over years.

Whether it helps you depends on where you were in the queue

It is worth being blunt about the distribution. If you watch rating actions, hold a direct plan and can place a redemption within the hour, segregation takes something from you: in the worked example you realise 9.52 instead of 9.80. That difference was funded by other unitholders in the same scheme, so it was never really yours, but it is real and pretending otherwise is dishonest.

If you are the other kind of investor, and most people in a debt scheme are, the arithmetic runs the other way: 9.52 instead of 9.10. That gain is not skill and not compensation. It is the removal of a penalty you were paying for being slower, which existed only because a scheme has to publish a price for something that does not have one.

What follows is a filter rather than a strategy. The question worth asking of a debt scheme is not whether it will ever side pocket, but whether it holds credit capable of producing a credit event at all, and if so whether its documents carry the enabling clause. Reading what you own before you need to is the same habit that makes overlap between funds and the disclosure regime of a discretionary managed account worth checking.

Three tools, three stages of the same problem

Segregation is one of a family of anti-run measures, staged rather than alternative, each acting at a different point on the same timeline.

The first two assume the assets have prices and that the problem is cost. Segregation is the tool for when that assumption fails, and the only one of the three that admits an asset cannot be priced at all.

What the mechanism is actually for

A segregated portfolio does not reduce a loss, accelerate a recovery, or improve anybody's outcome in aggregate. Every rupee it protects one investor from is a rupee another would otherwise have taken. Its function is to make the loss land on the units that were exposed to that credit when it happened, and not on whoever was slowest to react to a rating notice.

That is a narrow claim, and narrow claims are the ones worth trusting. The failure it prevents has happened at scale in Indian debt funds, and it is one of the few in pooled investing where the slow investor's loss is almost exactly the fast investor's gain. Once you see the transfer, the language of freezing reads differently: the money was never liquid, only the assumption that it was, and whoever moved first was cashing that assumption out.

The useful work sits upstream, in knowing what you own inside a pooled vehicle and what its documents permit under stress, whether the wrapper is a mutual fund or anything else. A mechanism built this carefully is a signal about the size of the problem it solves, and the problem is not the default. It is the queue.