How to Organise the Records a Transaction Leaves

Filing a fund transaction well means holding four items together: the instruction given, with whatever came back acknowledging it, the figure per unit and the day it sits on, the unit count actually booked, and any deduction together with the base underneath it. Without the day the transaction stops being checkable, however tidy the rest looks.

Almost nobody files badly on purpose. Paper arrives one piece at a time instead, each piece is put somewhere sensible on the day it lands, and the sensible places turn out to be different places. A year later a figure looks odd, the file is opened, and the parts needed are in four locations with nothing tying them to each other. The test of a set of records is not whether it looks organised but whether one transaction, picked without warning, can be walked from end to end out of it.

The scheme used throughout is the Girnar Large Cap Equity Fund, run by Girnar Asset Management Limited and open ended in form. Its net assets stand at Rs 4,200 crore, 120.00 crore units are in issue, and dividing the first by the second puts one unit at Rs 35.00. Kalyani Bhagat runs the portfolio; operations sit with Sohail Merchant. Five more appear as roles only, with no names attached: distributor, custodian, auditor, trustee company, and registrar and transfer agent.

Three matters are covered separately, and this guide leans on them. A unit is a slice of the scheme, and what one slice is worth gets struck inside the books the scheme keeps. Which struck figure an application gets considered against turns on conditions the Securities and Exchange Board of India (SEBI) writes. And any transaction can be put through a fixed run of checks, covered separately. Evidence is what those checks run on, and arranging it is a discipline of its own.

What is a fund record actually for?

The answer governs every later choice about filing. A record earns its keep only when a stranger, arriving afterwards with none of the context, can take one transaction and re-run it. Nine times in ten that stranger is the holder, a year on, with no memory of the day at all.

Re-running one transaction is a far harder standard than having the papers, and most filing passes the tidiness test while failing that one outright. Having the papers means the paper exists somewhere. Being able to recheck means that a specific question, asked cold, can be walked to an answer using only what is in the file, without any of the context that lived in one person's head on the day. Having the papers and being able to recheck are not close to the same thing, and the gap between them is where almost every filing problem lives.

Think about how a household keeps a receipt for a large appliance. The receipt goes in a drawer. Two years later something goes wrong and the question is whether the item is still covered, so what is needed is the receipt, the date, the model actually delivered and the terms of the cover, and the drawer has the receipt. On the day, everything else was obvious. Nobody wrote any of it down. A record setEverything kept about one occasion, held together as one thing rather than split up by what kind of paper each piece happens to be. is the fix for exactly that: it treats an occasion as the thing being filed, not a piece of paper.

Once a record is understood to exist for rechecking by a stranger, three consequences follow without argument. A missing input stops the walk rather than slowing it, so the set has to be complete. The question always arrives shaped as one occasion, so the set is assembled per occasion. And a restatement is one more thing that has to be trusted, so the set holds originals rather than the holder's own restatements of them.

Which four items must every transaction leave behind?

Four of them, working as a set and not as a wish list. The instruction given, and whatever came back acknowledging it. The value per unit, together with the day it sits on. The unit count that actually went onto the holding. Any deduction, and the base underneath it, written out.

Treat those as four slots that every transaction fills, and treat an empty slot as an answer rather than as a gap. A purchase that carried no deduction fills the fourth slot with the words no deduction. The two written words are information. A slot left blank because nobody looked is not information, and six months later the two are indistinguishable.

Lose one of the four and the transaction stops being checkable. The day underneath the value per unit reads as formatting and behaves as an input, and the day is the one that goes. A unit count feels like the answer, so holders keep it. Holders wrote the instruction themselves, so they keep that too. The deduction irritated somebody at the time, so it stays as well. The day sitting under a figure of Rs 35.00 disappears, and once it has, there is no route to establishing which struck figure Rs 35.00 even was.

Slot one takes the instruction and its acknowledgementWhatever came back saying an instruction had been received, which is not the same thing as the later document reporting what that instruction produced., and those two count as one item rather than as two. The thing asked for sits in the first, and confirmation that the asking landed sits in the second. Held side by side, they separate a case where nothing reached the scheme at all from a case where something reached it and was then handled differently from the holder's recollection.

Slot two takes the value per unit with its day. Here that figure falls out of the scheme's own books: Rs 4,200 crore of net assets carved into 120.00 crore units, giving Rs 35.00 for a single unit, and the division rather than the answer is what the file should hold. Slot three takes the unit count booked against the folioThe reference a scheme books a holder's units and details against. Folios are set out in full separately.. The booked count is the scheme's number and not the holder's own arithmetic. Rupees standing alone check nothing at all, so slot four takes any deduction with its base, in that order.

Why organise by transaction rather than by kind of paper?

Because the question eventually asked is about a transaction, and never about a kind of paper. Nobody has ever walked to a cupboard wanting to see all their confirmations. Holders want to see what happened on one occasion, and they want all four items of it at once.

So the organising axis runs folio first, then scheme and plan, then date. Folio first because that is the account the entries sit against. Scheme and plan next because the same holder can be in two schemes and in two plans of one scheme, and figures from different plans are not comparable with each other. Date last. Within one scheme in one folio, the only thing left to separate two entries is when they happened.

Sorting by kind of paper takes one transaction and puts its four items into four different places. The result looks orderly and makes the checking motion impossible. Every confirmation together, every statement together, every scheme document together. Each pile is internally consistent. Not one of them points at the other three, so reassembling a single occasion means four separate searches plus a judgement in each pile about which piece of paper belongs to the occasion being chased.

A pharmacy is the everyday version. Sort the shelves by tablet colour and the shop looks magnificent, and every prescription becomes four trips to four shelves with a guess at each one. Sort by what gets dispensed together and the shop looks duller and works. Storage that fights the retrieval question is not storage, it is decoration.

Why is a standing instruction kept apart from the lines it produces?

Because one registration is one document and the transactions it produces are many, each priced on its own occasion. The registration says what should happen repeatedly. The registration was written before any particular occasion, so it says nothing whatever about what happened on one.

A standing instructionOne registration given once, which then produces a separate transaction on each occasion it runs, every one priced on its own. registered for Rs 10,000/- an occasion into the Girnar Large Cap Equity Fund is a single sheet. Three occasions later there are three transactions, at values per unit of Rs 35.00 and, in this illustration, assumed values of Rs 32.00 and Rs 38.00. The three occasions produce 285.714 units, 312.500 units and 263.158 units, adding up to 861.372 units for Rs 30,000/- put in. Not one of those five numbers is written anywhere on the registration.

Hold the registration by itself and not one of the transactions it threw off can be rechecked. The hole turns up most often in a file that is otherwise careful. The reason is easy to see. The registration was signed and it took effort, so it feels like the important document. Each instalmentOne occasion of a standing instruction running, treated as its own transaction because it is priced and recorded on its own. afterwards arrives quietly, looks identical to the last one, and gets treated as routine confirmation of something already known.

The everyday parallel is a monthly milk arrangement with a vendor. The arrangement is one conversation. The deliveries are many. If the rate changed at some point during the year, the arrangement was agreed before the change and will not say when. Only the delivery records carry that. Same shape, same lesson: a registration is a rule, and the lines it produces are facts.

Why must the scheme's own documents sit beside the transactions?

Because a deduction is only checkable against a stated term, and stated terms live in the scheme's own documents rather than in any statement. A statement can show that Rs 350.00 came off. A statement cannot show why that was the right figure.

Take the redemption already worked through above. A holder redeems 1,000.000 units, priced at Rs 35.00 apiece, from the Girnar Large Cap Equity Fund. The redeemed amount is Rs 35,000.00. An assumed charge of 1.00 per cent against that amount removes Rs 350.00 and Rs 34,650.00 goes out. The 1.00 per cent is an assumed rate. A scheme sets its own rate in its own documents, and a different rate would move the Rs 350.00 in direct proportion.

Where a deduction cannot be traced back to a stated term, it cannot be checked at all, no matter how plainly the statement prints it. By itself, Rs 350.00 is a result. Testing it takes two more things. The first is the base it was applied against, the redeemed amount of Rs 35,000.00 rather than the money paid out or the units involved. The second is the load termsConditions a scheme sets down in its own documents for a charge on units coming out. An authority fixing a rate is a different thing. in the scheme's documents that a charge shaped like that comes from.

The second requirement has consequences for the filing. Scheme documents are not correspondence and never arrive with a holder's name on them, and that is precisely why they finish up nowhere. Scheme documents are the only thing in the file that says why a number was permitted to be what it was, so they belong with the transactions.

What does the filing look like once these transactions are in it?

Here it is, built rather than described. The transactions are the ones worked through above, and they are filed the way the four items ask for. The figures marked assumed stand in for values the scheme's books would supply, and the arithmetic runs the same way whatever those values turn out to be.

Two of those cells repay a closer look. Nowhere is Rs 35.00 simply quoted: it is what Rs 4,200 crore of net assets comes to across 120.00 crore units, and the file should carry that division and not just its answer. The unit count in the first transaction is also not a clean number. Rs 1,00,000/- against Rs 35.00 works out at twenty thousand sevenths of a unit, running 2,857.142857143 without ever stopping, so the booked 2,857.143 stands one part in seven thousand of a unit above the true figure. The booked figure pushed back through Rs 35.00 lands on Rs 1,00,000.005, precisely midway between two paise. Nothing settles that without a stated rounding rule.

Now file the standing instruction. A careful set of records usually gives way at exactly this point. A single registration at Rs 10,000/- each occasion threw off three transactions. On the first, the figure of Rs 35.00 comes from the record; the Rs 32.00 and the Rs 38.00 behind the other two are assumed here.

The residues carry signs and they do not cancel. One rounded down, one is exact, one rounded up, and the three together leave the recorded total 0.000180451 of a unit below the exact one. The shortfall is a genuinely tiny amount and it is still worth writing down. A residue nobody has looked at is indistinguishable from an error nobody has found.

Summing 285.714, 312.500 and 263.158 into 861.372 consumes only the three figures under examination and will agree with itself whatever they say. Something incapable of failing is not a check. The route that can fail runs the other way. Multiply each recorded unit count by the value per unit it was recorded against and compare the total against the money that actually left the bank account. The three products are Rs 9,999.99, Rs 10,000.00 and Rs 10,000.004. Together they come to Rs 29,999.994 against Rs 30,000/- put in, a residue of minus Rs 0.006. The three unit residues have been carried back into rupees at their own prices, and they do not cancel either.

The rupee check can fail, and that is the whole point of it. If a unit count on one occasion had been recorded against a different value per unit from the one filed beside it, the rupee total would land somewhere other than a fraction of a paisa away from Rs 30,000/-, and that would show. The addition never could have shown it.

One more figure comes out of the same rows. The average cost per unitMoney put in, over units received. Rebuilding it takes every occasion, and no closing figure will hand it back. is Rs 30,000/- over 861.372 units, or Rs 34.8282. Note what it took to get there: three amounts, three values per unit and three recorded unit counts, all of them filed. Take away any one occasion and the figure is not approximately right, it is unavailable.

How can the filing be tested?

A transaction chosen without looking, then walked from one end to the other. Not the one remembered clearly, and not the largest. A random one. The whole point is to find the gap nobody knew about rather than to confirm the part already trusted.

Run on the middle occasion of the standing instruction: Rs 10,000/- over an assumed Rs 32.00 is 312.500 units, and 312.500 units is what was recorded, so it ties exactly. The walk takes about ten seconds and it establishes that the file works for that transaction.

Now do the more useful half of the test. Break the record set deliberately. Take the day away from the value per unit and try the same walk again. The walk does not get harder, it stops before it starts. There is no longer any way to establish which struck figure Rs 32.00 was, and every later step depends on it. Nothing else has gone: the instruction, the amount, the value per unit itself and the recorded unit count are all still sitting there.

A set of records that cannot support a transaction picked at random has already failed, and finding that out on purpose costs seconds while finding it out by accident costs a great deal more. The reason to pick at random is that memory is not evenly distributed. Memory keeps the large transaction, the one that went wrong, the one that was argued about. Exactly those transactions have their gaps filled from memory without anybody noticing it happening.

What should be checked the moment a document arrives?

The document goes against the instruction actually given, while the giving of it is still fresh. Comparing the two is the cheapest check available anywhere in the whole business of holding units, and it is cheap only in the short window before the instruction fades.

The document is read against the instruction actually given rather than against what was expected to happen: does the amount or the unit count it acted on match what was asked for, is it the right folio and the right plan, does it carry a value per unit with a day written beside it, and if anything was deducted does it name the base that deduction was struck on. Four questions, one pass, and every one of them maps to one of the four slots.

Then the part everybody skips. Keep whatever acknowledged the instruction in the same place as the document that reports what the instruction produced. The acknowledgement feels superseded the moment the outcome arrives, so it is the item that goes missing first. Nothing has superseded it. The acknowledgement is the only thing in the file that establishes what was asked for, as distinct from what happened.

The list leaves out three things as well. How quickly a document should reach a holder, how often anything is sent and what a document has to contain are all set rather than chosen, and all three are read at sebi.gov.in. The internal question survives without any of them: does this document match the instruction it claims to be answering.

What follows when one of the four items is missing?

Go to whoever holds the registerThe maintained record of who holds how many units in a scheme, kept by one party rather than by everyone who touches a transaction., bearing in mind that only the party maintaining a record can alter it. For a scheme, the holdings register is kept by the registrar and transfer agent acting for the asset manager, so a question about an entry belongs there and nowhere else.

A distributor cannot change a record. Knowing that beforehand saves a whole round of correspondence with the wrong party. A distributor may have placed the instruction and may well help raise the query, and both are useful. A distributor does not have the register itself. Neither does the auditor, who examines the accounts of the scheme rather than the entries on a folio, and neither does the trustee company, whose job sits at a different level entirely.

If the units went into a demat account instead of straight onto a folio held with the scheme, the entry lives on a depository record and the question goes somewhere else entirely. The depository side is set out by the National Securities Depository Limited (NSDL) and by Central Depository Services (India) Limited (CDSL), whose sites are nsdl.co.in and cdslindia.com.

The query is then raised in the shape the four slots give it: naming the folio, the scheme and the plan, naming the transaction by the day it is believed to have happened, saying which of the four items are held and which one is not, and asking for the missing one specifically. A query that names three known items and one gap gets a precise answer. A query that says an amount looked wrong gets a general one, and several rounds later the problem is still being described.

Whatever comes back then goes into the same record set as the transaction it belongs to, not into a correspondence pile. The whole discipline sits in that one move. The answer to a question about a transaction is part of that transaction's evidence, and it is worth nothing filed anywhere else.

Why must a holder's own summary never stand as the source of truth?

Because a figure written down by hand, once it is separated from the document it came out of, is a figure nobody can check, including the person who wrote it. A summary is a finding aid. The moment it becomes the only place a number lives, it has quietly turned into the record itself.

As an index a summary earns its place; as an original it wrecks the file, and the only thing separating the two is whether the document behind it survives. A single line noting which transactions live in which place is genuinely handy and costs nothing. Once the documents behind that line are gone, the line has become the file, and the file is now one person's transcription and nothing else.

Make it concrete. Say the holder had jotted down about 861 units for Rs 30,000/- and held on to nothing else from the standing instruction. A rounded jotting now stands as the source of truthThe original a figure came off, as opposed to any later note about that figure. . Rs 30,000/- over 861 units is Rs 34.8432. Rs 30,000/- over the actual 861.372 units is Rs 34.8282. The two figures differ in the second decimal, and the note gives no way to tell which is right. The three occasions behind it are simply gone.

The same logic runs the other way for a balance, the most trusted document in the whole file. Take the Girnar Large Cap Equity Fund. Its closing figure of 1,857.143 units is arithmetically flawless and says nothing whatever about the route it took. Ask it for the figure per unit either transaction went through, for the day either of those figures sits on, for the Rs 350.00 removed or the base underneath it, and there is nothing there. Each of those was an input, and results do not carry inputs.

Which questions will a well kept record set still not answer?

Some, and knowing which ones spares a hunt for answers that were never in the file to begin with. Chief among them: did the correct day's figure attach to the transaction at all. The answer turns on conditions SEBI writes, not on anything a holder keeps.

A file can establish four things about a transaction, and it establishes them well: the value per unit applied, the day that value was said to belong to, the unit count that went onto the folio, and what came off with what base. Every one of those is a fact about the holder's own transaction, and every one of them is inside the four slots.

Well kept records get a question into a shape somebody can be asked; they do not answer it, and that split of work is right rather than a defect in the file. Which struck figure ought to have attached hangs on conditions about the arrival of an application and of the money. SEBI makes those, sebi.gov.in publishes them, and they get read there on the day rather than recalled. No amount of care on the holder's side turns that into something settled from a cupboard.

Knowing the boundary marks when to stop searching. A holder who does not know the boundary keeps turning over paper looking for an answer that was never going to be in it, concludes their filing must be deficient, and either gives up or files even more of the wrong thing. The boundary converts a vague unease into a specific query put to a specific party, and that conversion is the whole return on the work.

Girnar Asset Management Limited, the Girnar Large Cap Equity Fund, Kalyani Bhagat and Sohail Merchant are invented.
Educational material. Not advice on any investment, tax, budget or market position.