Reconciliation: Matching a Scheme's Units, Cash and Holdings

Three ties run through a scheme every working day. Units listed on the register are set against the unit capital line in the books. Money sitting at the bank is set against the cash line. Securities held outside by the custodian are set against the holdings line. Separate parties write the two sides of each tie, and that separation is the whole reason agreement counts for anything.

The word itself trips people first. Tidying is what it sounds like, and the work is close to the opposite of tidying. ReconciliationSetting two separately written records side by side and accounting for every rupee of whatever gap appears between them. means accounting for every rupee of the gap between two records that were built apart from each other. Making the gap disappear is not the target, and on a bad day it is the damage. A difference that has been explained is finished work. A difference that has merely been removed is a hole with a lid on it.

Everything worked below runs off one invented scheme. Girnar Asset Management Limited is the manager, and the scheme it runs here, the Girnar Large Cap Equity Fund, is open ended and holds equity. Put its net assetsEverything a scheme has, once everything it owes has been taken off. Net assets are the top half of the value per unit. of Rs 4,200 crore over its 120.00 crore units in issue, or 1,20,00,00,000 of them spelled out, and a unit comes to exactly Rs 35.00. Standing behind the total are 3,80,000 folios. Sohail Merchant heads operations and Kalyani Bhagat runs the portfolio.

Four matters are settled elsewhere. The register, the custodian's own record and the scheme's books are each covered separately, as is the arithmetic that turns a rupee amount into a carried unit figure. So is who may change what: the registrar and transfer agent writes the register and nothing else, the custodian moves securities and writes nothing on the register, and the operations team writes the valuation record and touches neither of the other two. Reconciliation begins where those separate records are laid against each other: why they can disagree at all, and what it means when one of them will not line up.

What exactly gets set against what?

Three pairs, and it is worth naming them slowly because everything after this uses them. The first pair is units. The register carries one line for every folio, and those 3,80,000 lines add to a total; the books carry a single figure called unit capitalThe line in a scheme's books carrying the total number of units in issue, held as one book figure rather than as a list of holders. that says how many units exist. The register total and the unit capital figure are answers to the same question written by different hands, so the two are set against each other.

The second pair is money. The scheme has a bank account with a balance in it, and the books have a cash line. The third pair is securities. The scheme's shares and bonds sit in custodyThe arrangement under which a separate party holds a scheme's securities and keeps its own record of what it is holding. with a record kept there of exactly what is held, and the books carry a holdings line saying the same thing in the scheme's own words. Set one against the other.

A street vendor runs the identical shape on a scale small enough to hold in the head. She has a cash box, a notebook where she writes every sale, and a cart with stock on it. At the end of the day she can compare the notebook against the cash box, and she can compare the notebook against what is left on the cart. If the notebook says fourteen plates went out and the cart is down by seventeen, something happened that the notebook does not know about. She is not tidying the notebook. She is finding out what happened.

Taken together the three pairs account for the whole of a scheme: whose it is, what it is holding, and how much of that is genuinely its own. Drop any one of them and a complete side of the scheme goes unchecked. Units alone would let the securities drift. Securities alone would let the register drift. Each pair covers a side the other two leave open, so all three run, and they run as three separate exercises rather than one.

Why does it matter who wrote each side?

Because a match is only evidence if the two sides could have disagreed for a reason worth knowing. A shopkeeper who counts the notes in her own cash box twice has checked her counting. Handing the box to somebody else to count checks the cash itself. The second exercise can fail in a way the first cannot, and that is the entire value of it.

Run the same test on records. Suppose one team pulls a figure out of one system and writes it into two different reports. The two reports agree. The agreement establishes one thing: nobody mistyped. Every assumption inside the source travelled into both reports untouched, so a fault in the source shows up identically on each side and cancels itself out of the comparison. A tie is worth precisely how far apart its two sides were built, and independenceTwo records counting as separate evidence, because neither was copied from the other and neither was built by the same hands from the same starting figure. is what a tie trades on. Where both sides came off one source through one pair of hands there is none of it to trade, and agreement establishes only that the copying was careful.

Independence is why the four roles in a scheme's operating chain are separated the way they are. The split looks at first like a division of labour, as though the work were simply too much for one office. Volume is not the main reason for it. The registrar and transfer agent, the custodian and the operations team hold three records that were never derived from one another. Nothing was copied from anything, so agreement between them means something. Split the work differently and the work would still get done; there would simply be nothing left to check it with.

Does forcing two numbers level count as reconciling?

No, and the gap between those two acts is the sharpest distinction in the whole of fund operations. Reconciling means accounting for the difference. Levelling means removing it. The two acts can produce the same pair of numbers at the end of the afternoon and leave behind completely different amounts of evidence.

The two routes leave different things behind. Explaining a difference means writing down what caused it, and afterwards anybody can re-read that explanation and test it. The trail survives. Where the answer is instead an adjustmentAn entry passed to move a figure in one record so that it lines up with another record. to bring one record onto the other, the two numbers agree, and what survives is an entry with an amount on it and no cause behind it. Driving a gap to nil with a posting that has no account behind it is worse than leaving the gap exactly where it sat. A gap left alone can still be looked at, and a gap driven out cannot.

The uncomfortable part is what that does to anybody looking later. An entry that quietly tidied away a harmless rounding leftover and an entry that covered a genuine shortfall look exactly alike once both are posted. Same shape, same amount, same absence of a reason. The one thing that could have separated them was the difference itself, sitting there unexplained and visible, and that is precisely what the entry removed. Seniority does not repair the loss either. An approval records who agreed rather than what the money was.

How is one kind of gap told from another?

By watching it, not by staring at it. A single day's difference says almost nothing about which kind it is. The behaviour across a run of strikes says nearly everything, and there are three kinds worth separating.

A timing differenceA gap that appears only because the two records were struck at different moments, and closes once the later one catches up. is real, entirely expected, and closes itself. The two records were written at different moments and one of them has something the other has not caught yet. Nobody is wrong. Wait for the later record to catch up and the gap goes to exactly zero without anybody touching it.

A rounding residualThe small leftover created when a figure is carried to a fixed number of decimals instead of its exact value. is real, tiny and permanent. A convention rather than a mistake created it, and a convention does not correct itself. Across a run of strikes the residual is exactly the same size on every one of them. The residual has to be tracked rather than ignored, and one is worked out below to the last fraction of a paise.

An error is neither expected nor self clearing. An error sits there, or grows as more of the same fault accumulates, and nothing in the arithmetic explains it. Somebody has to go and find it. A gap followed over several strikes declares itself: timing takes itself down to nil, a rounding leftover holds its size without budging at all, and an error does neither of those for any cause that could be set down on paper.

The difference itself has a name. A breakA gap between the two sides of a match that has not yet been explained. A break is a question waiting for an answer, not a verdict on anybody. is what an operations team calls a gap it has not explained yet, and the word carries no accusation with it. A break is not automatically anybody's fault: it is a difference waiting to be explained, and the explanation is at least as likely to be a rounding convention, or the two records having been struck at different moments, as it is to be a mistake by anyone. Treating every break as a hunt for a culprit is how teams end up passing entries instead of asking questions.

What is left over when a unit figure is rounded?

Something exact, small, and permanent, and this record can produce it to the last fraction of a paise. Take a purchase of Rs 1,00,000/- into the Girnar Large Cap Equity Fund at a value of Rs 35.00 a unit. Divide. One lakh over thirty five is twenty thousand sevenths of a unit, and twenty thousand sevenths does not terminate: it runs 2,857.142857 and keeps going forever.

Nobody can write forever onto a register, so the figure is carried to a fixed number of decimals. Take three, and the carried figure is 2,857.143 units. The carried figure is above the exact quotient, not below it. By how much, exactly? By one seven thousandth of a unit. Do the check rather than believe it: one hundred and forty three thousandths minus one seventh is a thousand and one over seven thousand, less a thousand over seven thousand. Subtracting leaves one over seven thousand. So the residual is exactly one seven thousandth of a unit. As a decimal the residual begins 0.000142857 and never ends either.

Now price it. One seven thousandth of a unit at Rs 35.00 is thirty five over seven thousand, and thirty five over seven thousand cancels to one two hundredth of a rupee. One two hundredth of a rupee is exactly Rs 0.005. Half a paise, exactly, not approximately. And it is not notional: redeem the carried 2,857.143 units at the same Rs 35.00 and the scheme pays out Rs 1,00,000.005 against the Rs 1,00,000/- that came in.

Read the sign on that last row. The sign is the teaching, and the sign is positive. The register side is larger than the money side, so the residual on this order runs one way rather than being as likely to fall either way. Rs 1,00,000.005 is also exactly half way between Rs 1,00,000.00 and Rs 1,00,000.01. Half way is the least convenient place a figure can land when somebody has to decide which of the two paise to write down. Where a convention carries a figure in one direction only, every residual it creates carries the same sign, so the residuals add up across orders rather than offsetting each other. A tiny permanent difference has to be tracked rather than waved through.

Two things behind that arithmetic are fixed outside the arithmetic. How many decimals a scheme carries a unit figure to, and which way it rounds, are set by the market regulator and are read at sebi.gov.in; three decimals and a carried up figure are the illustration used above. The second is the number of orders a scheme takes in a day. No rule fixes that count, and it moves from scheme to scheme and from day to day. So the residual is shown on one order and never multiplied out to a day's worth.

What does it look like when all three hold?

Laid out on the Girnar Large Cap Equity Fund, at rest, with nothing wrong anywhere. The register's 3,80,000 folio lines add to 1,20,00,00,000 units, and the unit capital line in the books says 1,20,00,00,000 units. Subtracting one from the other gives exactly zero. Zero is a real result and not a blank. Net assets of Rs 4,200 crore, written out as Rs 42,00,00,00,000, divided by those units is Rs 35.00 a unit exactly, and those same net assets have to be supported by what sits in custody, plus what sits at the bank, plus everything the scheme is owed, less everything it owes.

A total is not a split. Rs 4,200 crore of net assets is one figure, and one figure carries no securities amount, no bank balance, no receivable and no payable inside it. So the third tie can be named and its shape drawn, and its two sides cannot be filled in from a total alone.

Now a trap that catches people who like checking things twice. Running a second check alongside the first is tempting: divide net assets by units to get the value per unit, then multiply units by the value per unit to get back to net assets, and treat the pair as two confirmations. The pair is not two checks. The two rows are one equation rearranged, and the units figure cancels out of the second: if the value per unit was made as net assets over units, then units multiplied by that value is units times net assets over units, the units cancel top and bottom, and net assets is all that is left, so the second row restates the first and can never fail while the first is how the value was built.

So where is a check that can actually fail? Sum the 3,80,000 lines on the register and set that total against the unit capital line. The register total draws on a record the other two never touched. The register was written folio by folio by the registrar and transfer agent as orders came in; the unit capital line was written in the books by the operations team. Neither was derived from the other, so the two can genuinely disagree, and when they agree, something has been established rather than restated. The custody tie has the same property for the same reason.

How far does an unexplained Rs 1.20 crore move a unit?

Suppose the three ties have run, the timing differences have cleared, the rounding residuals are accounted for, and Rs 1.20 crore is left over that nobody can account for anywhere. Written out, that is Rs 1,20,00,000/-. Here is the same number read four ways, with every division shown.

Against the units first: Rs 1,20,00,000/- divided by 1,20,00,00,000 units is Rs 0.01, exactly one paise, with no rounding involved anywhere. So a value that ought to be Rs 35.00 is being published as Rs 35.01 or as Rs 34.99, depending on which way the difference runs. Against the size of the scheme second: Rs 1,20,00,000/- over Rs 42,00,00,00,000 of net assets is 0.02857 per cent, a figure so small that it disappears into any sentence it is put in.

Against what the scheme spends third, and that reading holds as a fraction of a year rather than as anything shorter. At 1.65 per cent, the expense ratio, against the Rs 4,200 crore of net assets, a year of running cost comes to Rs 69.30 crore, written out as Rs 69,30,00,000/-. Rs 1,20,00,000/- over Rs 69,30,00,000/- is 1.7316 per cent of a whole year of everything the scheme spends. Put into a shorter unit on a 365 day year, the same amount is about 6.32 days of running cost. A 365 day year is an assumption of this illustration and not a convention any rule fixes. Against one holder fourth: Rs 1,20,00,000/- across 3,80,000 folios is about Rs 31.58. The average folio of about 3,157.89 units multiplied by that one paise comes to the same Rs 31.58.

The same number is too small for any holder to question and larger than one and a half per cent of everything the scheme spends across a whole year, and both readings are true at once, with neither cancelling the other. The first reading is why nobody outside notices. The second is why nobody inside gets to shrug.

One thing this does not do, and it needs saying plainly because the arithmetic invites the opposite conclusion. None of this protects anybody from losing money. Three records agreeing establishes that the count is right; it establishes nothing whatsoever about whether the count is worth anything. A scheme whose holdings fall steadily in value can reconcile perfectly on every tie, every strike, all the way down. Reconciliation is about whether the number is the number, not about whether the number is good.

Why does this never reach anybody outside?

Because of what happens to it on the way out. The difference is divided twice, and each division is by a large number. The first division is by the unit count, and that division takes Rs 1,20,00,000/- down to one paise. The second happens when a holder opens a statement. A statement shows one folio, and one folio is a share of the whole. The average folio here runs about 3,157.89 units, so one paise on each of them is about Rs 31.58.

Nobody queries Rs 31.58. Holders would care; there is nothing on the document to query. A statement shows a unit count and a value per unit, and the gap is buried inside the value per unit rather than sitting next to it. Nothing on the statement says anything is out. By the time a gap gets to a holder it has been through two divisions, the unit count first and then whatever slice one folio represents, so its full size stays visible to the three parties running the checks and to nobody else.

None of that criticises holders, and none of it is a defect in the statement. The shrinking is structural, and the shrinking is the reason the checking is arranged the way it is. The people with a chance of noticing Rs 1,20,00,000/- are the ones who see it as Rs 1,20,00,000/-: the registrar and transfer agent, the custodian and the operations team, each looking at their own record before it is divided by anything. Pushing the checking outwards to the people holding units pushes it to the exact point where the number has been shrunk past visibility.

Who fixes the timing, and why is none of it here?

Timing falls to the market regulator, not to the asset manager. The interval at which each pairing has to be run, the duty that arises where a gap refuses to explain, who that duty runs to and the deadline hanging off it are all the market regulator's to fix, and every one of them is open to revision. Whatever the position is on any given day sits at sebi.gov.in.

The whole of the mechanism above survives an interval being changed: three pairings, the demand that each side come from separate hands, three behaviours that separate one kind of gap from another, one division down into the published value of a unit, and the rule that an account rather than a posting is what shuts a gap. A change in the required frequency tomorrow would leave not one sentence of that needing to be rewritten. The mechanism above this line holds because it is written without a single number that a regulator sets.

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