You were not measured at the close, and that is the whole of the mechanism
The short answer
Margin adequacy is read from unannounced snapshots taken during the session, not from the state of your account at the close. A shortfall is the requirement at a sampled moment less the margin available at that moment, and it is then read both as a rupee amount and as a proportion of the requirement it came from, with the worse reading placing it in a band. Repetition inside a defined window is priced separately from size. In the illustrative day computed on this page an account that was funded at the open and funded at the close was short for 38 minutes in between, by as much as 2,40,000. At four snapshots, an illustrative count rather than a stated one, the chance of the excursion being caught on any single day is about 35 per cent, and across a month of 20 sessions it is 99.98 per cent. The practical rule that follows is that the peak requirement is what has to be funded, never the closing one.
This page states no penalty rate, no rupee threshold and no band percentage. That is deliberate and the reason is given in full in the Sources block: those figures are set by the exchanges and the clearing corporations, they are revised, and they could not be verified at the time of writing. Everything else here is mechanism, and the mechanism is what people actually get wrong. The arithmetic on this page is computed from inputs stated alongside it, and every figure in it is illustrative.
The close is not when you were measured
Almost every account that incurs this charge does so believing it was never short of margin, and the belief is sincere. It rests on a picture of margin as a property of a trade: you had enough when you opened the position, you had enough when you closed it, and the ledger at the end of the day shows a comfortable balance. Under that picture a shortfall looks like an accounting error.
The picture is wrong in one specific way. Margin is not a property of a trade. It is a property of the book, recomputed continuously, and it is compared against your available margin at moments you do not choose and are not told about in advance. A day in which the requirement was covered at the open, covered at the close, and uncovered for twenty minutes at noon is a day with a shortfall. The two states you saw were not sampled. The one you did not see was.
Everything else follows from that single change of measurement point. The size of the charge, the band it falls into, the way repetition is treated and the funding rule that avoids it are all consequences of the fact that the measurement happens where you are not looking.
Why the measurement moved off the clock
It is tempting to read intraday sampling as an extra rule layered on top of an older one. It is closer to the opposite: it is the removal of an exemption that the older design created by accident.
A margin requirement enforced only at the end of the day is enforced at a moment the participant controls completely. A position can be carried through the entire session on a fraction of its requirement and then closed before the measurement, and the account presents a compliant book because the thing that needed margin no longer exists. The requirement in that design binds only the positions nobody bothered to unwind. Everything intraday is free, and the risk that the requirement exists to cover, which is the risk of the market moving against an open position, is exactly the risk that the design fails to price.
You can see this in the shape of what the reform actually changed. Overnight requirements were not the problem, because an overnight position is by definition still open when the measurement happens. The gap was entirely intraday, and it was not a small one. Removing it did not require a new prohibition. It required measuring at moments that cannot be anticipated, which converts an obligation that was optional in practice into one that is continuous in practice.
The consequence for a trader is the part that is easy to state and hard to internalise. There is no safe moment. A requirement that stands for twelve minutes is as real as one that stands for six hours, and the only difference between them is the probability of being sampled while it stands, which is a matter of arithmetic rather than of luck.
A day with no shortfall at any moment you looked at it
Take one illustrative account through one session. The book is not exotic and the day is not dramatic. The account holds 3,20,000 of available margin throughout, and never adds to it or draws from it.
| Stretch of the session | Requirement | Minutes | What changed | Against 3,20,000 |
|---|---|---|---|---|
| 09:15 to 10:05 | 2,40,000 | 50 | opening position, hedged | covered |
| 10:05 to 10:17 | 5,60,000 | 12 | second leg entered before its hedge filled | short by 2,40,000 |
| 10:17 to 12:40 | 2,70,000 | 143 | hedge complete, requirement settles | covered |
| 12:40 to 13:06 | 4,80,000 | 26 | protective leg closed first | short by 1,60,000 |
| 13:06 to 15:20 | 1,60,000 | 134 | remaining leg closed | covered |
| 15:20 to 15:30 | 1,95,000 | 10 | small overnight position carried | covered |
Two things in that table are worth reading slowly. The first is the twelve minute stretch at 5,60,000. Nothing went wrong in the market during it. A second leg was entered before the leg that was meant to hedge it filled, so for twelve minutes the book contained an unhedged position whose requirement was more than double what the completed structure would need. The requirement was correct. The book really was that exposed for twelve minutes.
The second is the stretch at 4,80,000 after midday, which arises from closing a structure in the wrong order. The protective leg was closed first to take a profit on it, leaving the remaining leg standing alone for twenty six minutes. Notional exposure fell. The requirement rose, because the offset that had been reducing it was gone. This is the single most reliable way to manufacture a shortfall out of a position that is being reduced, and it catches people precisely because reducing a position feels like the safe direction.
Across the whole session the peak requirement was 5,60,000, the average was about 2,48,533, and the closing requirement was 1,95,000. Those three numbers differ by a factor of nearly three, and only one of them is the number that had to be funded.
| If the account had funded to | Level | Minutes short | Largest shortfall | As a share of the requirement then |
|---|---|---|---|---|
| The closing requirement | 1,95,000 | 231 | 3,65,000 | 65.2% |
| The average requirement across the day | 2,48,533 | 181 | 3,11,467 | 55.6% |
| What this account actually held | 3,20,000 | 38 | 2,40,000 | 42.9% |
| The peak requirement | 5,60,000 | 0 | 0 | 0.0% |
Funding to the closing requirement, which is the policy almost everyone follows without naming it, leaves the account short for 231 of the 375 minutes in the session. Funding to the average leaves it short for 181. Funding to the peak leaves it short for none, which is the only column in that table with a zero in it, and it is the only policy that is a policy rather than a hope.
The snapshot is a sampling problem, and the month is the real horizon
The obvious objection to all of this is that a short window is unlikely to be caught. That objection is arithmetically correct for a single day and badly wrong for a month, and the gap between the two is where most of the surprise lives.
If the requirement stands above the available margin for m minutes of a session of 375 minutes, and the clearing corporation takes k snapshots at moments drawn independently across the session, the chance that at least one of them lands inside the window is one minus the chance that every one of them misses. That is a single line of arithmetic and it has no free parameters beyond m and k.
| Minutes short | k = 1 | k = 2 | k = 4 | k = 8 | k = 20 |
|---|---|---|---|---|---|
| 5 minutes | 1.3% | 2.6% | 5.2% | 10.2% | 23.5% |
| 12 minutes | 3.2% | 6.3% | 12.2% | 22.9% | 47.8% |
| 26 minutes | 6.9% | 13.4% | 25.0% | 43.7% | 76.2% |
| 38 minutes | 10.1% | 19.2% | 34.8% | 57.5% | 88.2% |
| 60 minutes | 16.0% | 29.4% | 50.2% | 75.2% | 96.9% |
| 120 minutes | 32.0% | 53.8% | 78.6% | 95.4% | 100.0% |
Read the 38 minute row, which is the illustrative day above. At an illustrative four snapshots the chance of being caught on that day is 34.8 per cent. Two days out of three, nothing happens and the trader learns that the behaviour is fine.
Now repeat the behaviour, because a habit is what produced the window in the first place. Over 20 sessions the expected number of instances is 6.96, and the chance of at least one is 99.98 per cent. The behaviour that was invisible on a daily view is a near certainty on a monthly one, and by the time it becomes visible it has generated enough instances to be read as repetition rather than as an accident.
This is the structural reason the charge feels arbitrary to the people who receive it. The feedback arrives late, it arrives in a lump, and it arrives for a practice that appeared to be working every day it was observed. The sampling did not change. The number of draws did.
What the requirement will accept, and what it will not
A shortfall has two inputs and most attention goes to the wrong one. The requirement is the input people study. The margin available is the input that quietly does the damage, because the number in a holdings statement and the number the requirement accepts are not the same number, and there are three separate reasons for that.
The first is the haircut. Securities pledged as collateral are credited at less than market value, by a percentage set for each security by the clearing corporation and revised as volatility and liquidity change. The credit is the market value less the haircut, so the amount of stock that has to be pledged to produce a given credit is always more than the credit itself.
| Haircut | Market value needed | Value absorbed by the haircut |
|---|---|---|
| 5% | 3,15,789 | 15,789 |
| 10% | 3,33,333 | 33,333 |
| 20% | 3,75,000 | 75,000 |
| 30% | 4,28,571 | 1,28,571 |
| 50% | 6,00,000 | 3,00,000 |
The second is the cash component. A stated proportion of the requirement has to be met in cash or in cash equivalents, and non cash collateral is accepted only for the remainder. The proportion is published by the clearing corporation. It is not stated on this page, and it is carried below as a parameter written c, because a wrong figure here would be worse than none.
The mechanism is what matters and it is precise. An account holding cash of K and haircut adjusted non cash collateral of H can support a requirement R only if two conditions hold at once: the total accepted value must cover it, so K plus H must be at least R; and the cash component must be covered in cash, so K must be at least c times R. The largest requirement the account can support is therefore the smaller of K divided by c and K plus H. When cash is the binding term, adding more non cash collateral changes nothing at all.
| Cash proportion c | Cap set by cash | Total accepted value | Largest requirement supportable | Accepted value left idle |
|---|---|---|---|---|
| 0.25 | 6,00,000 | 6,30,000 | 6,00,000 | 30,000 |
| 0.40 | 3,75,000 | 6,30,000 | 3,75,000 | 2,55,000 |
| 0.50 | 3,00,000 | 6,30,000 | 3,00,000 | 3,30,000 |
| 0.60 | 2,50,000 | 6,30,000 | 2,50,000 | 3,80,000 |
Every row in which the cap set by cash is the smaller number is a row in which the account is carrying collateral it cannot use. At c of 0.40 this illustrative account holds 6,30,000 of accepted value, can support a requirement of only 3,75,000, and against the peak requirement of 5,60,000 from the session above is short by 1,85,000. Nothing about that shortfall is visible in a statement that shows total collateral against total requirement, which is the view most terminals present.
The third reason is timing. Funds transferred during the session are recognised by the clearing corporation on its own timetable, which is not the timetable on which a terminal updates a balance. An account that looks funded to its owner at the moment of the transfer may not be funded in the record that the measurement reads. This is why a transfer made in response to a margin call is a repair rather than a prevention: it fixes the state going forward and does nothing about the sampled moment that has already passed.
A small shortfall on a small account is not a small shortfall
Once a shortfall exists, its size decides how it is treated, and size is measured twice. The rupee amount is one reading. The amount as a proportion of the requirement it came from is the other. The schedule is banded, and the band is read from the worse of the two.
| Requirement at the sampled moment | Shortfall | As a proportion | How it reads |
|---|---|---|---|
| 1,20,000 | 18,000 | 15.00% | fails a proportional test long before an absolute one |
| 2,40,000 | 38,000 | 15.83% | fails a proportional test long before an absolute one |
| 9,60,000 | 38,000 | 3.96% | reads small on proportion, whatever the rupees |
| 24,00,000 | 38,000 | 1.58% | reads small on proportion, whatever the rupees |
| 24,00,000 | 2,40,000 | 10.00% | fails a proportional test long before an absolute one |
| 56,00,000 | 2,40,000 | 4.29% | reads small on proportion, whatever the rupees |
The asymmetry in that table is the part worth carrying away. A shortfall of 38,000 is 15.83 per cent of a requirement of 2,40,000 and 1.58 per cent of a requirement of 24,00,000. The rupees are identical. The proportional reading differs by a factor of ten, and it is the proportional reading that places the small account in the worse band.
This is not an accident of drafting. A proportional test is the only test that means the same thing on a book of two lakh and a book of two crore. An absolute test alone would make small accounts effectively unregulated and large ones charged for trivia, so both tests exist and the worse of the two governs. The practical effect is that a small account has very little room: an amount that a larger book would not notice is, on a small requirement, a significant fraction of it.
Repetition is priced separately from size
The second dimension of the schedule is frequency. A single instance is treated as an accident. Instances that keep recurring inside a defined window are treated as a funding practice, and they are charged at a multiple of the base rate rather than at it.
The multiplier is published. It is not stated here, for the same reason as everything else in that family. What can be stated, because it is arithmetic, is what a multiplier of any size does to a month.
| If the escalated band is | Month costs, in base charges | Average charge per instance | Excess over a flat month |
|---|---|---|---|
| 2 times the base rate | 13 | 1.62 | 62.5% |
| 5 times the base rate | 28 | 3.50 | 250.0% |
| 10 times the base rate | 53 | 6.62 | 562.5% |
| 20 times the base rate | 103 | 12.88 | 1187.5% |
The shape is the lesson. Because only the instances beyond the threshold escalate, the cost of a month is not linear in the number of instances and it is not linear in the multiplier either. It is flat up to wherever the threshold sits and steep after it, which means that at an assumed threshold of 3, the difference between 3 instances in a month and 8 is not 5 instances worth of charge. At a multiplier of ten it is 50 base charges on top of 3, an average of 6.62 base charges per instance across the month against a flat expectation of one.
This is also why the arithmetic of the previous section matters so much. A practice with a 35 per cent daily catch rate does not produce one instance a month. It produces close to seven, on these inputs 6.96, which is comfortably past any threshold an escalation is likely to sit at, and it produces them from behaviour the trader has no reason to think is unusual.
Fund the peak, not the close
Everything above collapses into one operational rule, and it is not a rule about discipline or about watching the screen more carefully. It is a rule about which number you size against.
The requirement that has to be funded is the highest requirement the book will reach at any moment during the session, including the moments when a structure is half built and the moments when it is half unwound. Sizing to the requirement the completed structure will carry is sizing to a number that is only true for part of the day. Sizing to the closing requirement is sizing to a number that is true for one instant, chosen by you, that nobody measures.
| Situation | Why the requirement spikes | What removes the spike |
|---|---|---|
| A hedged structure entered leg by leg | Until the offsetting leg exists, the book contains a naked position and is margined as one | Enter the leg that reduces the requirement first, or enter the structure as one order where the member supports it |
| A structure closed leg by leg | Closing the protective leg first removes the offset while the risk leg remains | Close the risk leg first, so the requirement only ever falls |
| A stop that converts a position | A triggered exit can leave a different net position with a different requirement for as long as it takes to fill | Size the account for the requirement of the post trigger book, not the pre trigger one |
| Collateral revalued during the session | Accepted value falls when prices fall or a haircut is revised, without the position changing at all | Hold a buffer in cash rather than at the edge of the cash component |
| Funds moved during the session | Recognition follows the clearing timetable, not the terminal balance | Fund before the session rather than during it |
There is a quieter point underneath the table. Every row in it is a decision taken before the order goes in. None of them can be fixed by attention after the fact, because the measurement has already happened by the time anything appears on a statement. That is unusual among trading costs, most of which are at least observable as they are incurred, and it is the reason this one is best treated as a sizing constraint rather than as a charge.
What this page will not tell you, and why
The specific rates, the rupee thresholds, the band boundaries, the number of snapshots taken each day, the cash proportion and the haircut schedule are all real, all published, and all absent from this page. They are set by the exchanges and the clearing corporations under the regulator's framework, they are revised as conditions change, and the detail a particular member applies when recovering the charge can differ in ways that matter to the number on your statement.
They could not be verified at the time of writing. The honest move when a figure cannot be checked is to omit it and say so, because a reader who is told to look something up loses five minutes, and a reader who is given a plausible wrong number sizes a position against it. Every quantity on this page is either computed from inputs printed alongside it or carried as a named parameter, and all of it is illustrative.
What does not move is the structure, and the structure is what this page is for: measured at unannounced moments inside the session, computed as the requirement less what was available at that moment, banded on the worse of an absolute and a proportional reading, escalated on repetition, and satisfied only by collateral of the right kind recognised at the right time. Read your own member's published schedule for the current figures, and read the circulars it cites rather than a summary of them.
Margin is not an administrative layer sitting on top of a strategy. It is a constraint that decides which strategies are available to a given account at a given size, and reading it as arithmetic rather than as paperwork is the difference between a position you can hold and one that is quietly costing you to hold it.
Frequently asked questions
Why is margin checked during the day rather than at the end of it?
Because an end of day check measures a moment the account controls. Anyone carrying an unfunded position through the session can square it off before the close and present a compliant book, so a day end measurement would have priced only the positions nobody bothered to unwind. Moving the measurement to unannounced moments inside the session does not add a rule. It removes the gap that made the old rule optional.
My position was funded when I opened it and funded when I closed it. How can there be a shortfall?
Because neither of those moments is when you were measured. The requirement is not a property of the trade you placed, it is a property of the book you held, and it moves whenever the book changes, whenever a hedge is incomplete, and whenever collateral is revalued. If it stood above your available margin at any sampled moment, that moment is the measurement, and the state before and after it is not relevant to the calculation.
How is the size of the shortfall worked out?
As the requirement at that moment less the margin available at that moment, taking only the positive difference. The number is then read against the requirement it came from, so it carries both an absolute size in rupees and a size as a proportion of what was required. The band is read from those two readings together, which is why the same rupee figure can be treated differently on two different accounts.
Why can a smaller shortfall be treated more severely than a larger one?
Because a proportion is one of the two readings. A shortfall of a given number of rupees against a small requirement is a large fraction of it, while the same rupees against a large requirement is a rounding error. Both an absolute test and a proportional test are applied, and the worse of the two readings is what places the instance. A small account can therefore be placed in a higher band on an amount a larger book would barely register.
Why does repetition cost more than the same total spread across separate accounts or months?
Because repetition is priced as a separate fact from size. The schedule is built so that a one off is charged at a base rate while instances that keep recurring inside a defined window are charged at a multiple of it, on the reasoning that a recurring shortfall is a funding practice rather than an accident. The escalation is what turns a series of small charges into a meaningful one, and it is the part most people discover only once it has already happened several times.
Does pledged collateral count toward the requirement?
It counts, after a haircut, and only up to a limit. The haircut reduces the credit given to the market value of what is pledged, so the value you see in a holdings statement is never the value the requirement accepts. Beyond that, a stated proportion of the requirement has to be met in cash or in cash equivalents, so non cash collateral can be plentiful and still leave the account short on the part of the requirement it is not allowed to cover.
Can I be short while holding more collateral than the requirement?
Yes, and it is one of the most common ways this is incurred. If the cash component is not satisfied, the account is short by the missing cash even though the total accepted value of everything pledged comfortably exceeds the requirement. The computed example on this page shows an account holding more accepted value than its peak requirement and still failing, because the surplus was all in the part that has a ceiling on it.
What are the actual penalty rates and thresholds?
They are set by the exchanges and the clearing corporations, they are published, and they are revised, so the only current figures worth having are the ones on your own clearing member's schedule and in the circulars it cites. This page deliberately states none of them, because the figures could not be verified at the time it was written and a confidently wrong rate is worse than an instruction to look it up. What does not change is the structure: banded by size, read against the requirement, and escalated on repetition.
Who pays the penalty, and does the broker keep it?
The charge is raised by the exchange on the clearing member or the trading member, and it is recovered from the client account whose shortfall produced it. It is not broker revenue. That matters for two reasons. It explains why a member will block or square off a position it considers under margined rather than wait for you to fund it, and it explains why the line item appears on your statement after the fact rather than as a rejection at the time of the order.
What actually stops this from recurring?
Funding to the peak requirement the book will reach rather than the requirement it will end on, and never creating a window in which an unhedged leg stands alone. Both of those are decisions taken before the order, not after the statement. Funds moved during the session are recognised on the clearing corporation's timetable rather than the terminal's, so an intraday transfer is a repair that may land after the moment it was needed.
What could not be verified for this page. No penalty rate, rupee threshold, band boundary, escalation multiplier, escalation threshold, cash component proportion, haircut percentage or snapshot count is stated anywhere above as a rule, because none of them could be checked at the time of writing: the search budget available to this page was exhausted and the regulator, exchange and clearing corporation sites were unreachable from where it was produced. Rather than repeat figures from memory, the schedule is taught in shape, and every rate sensitive quantity is carried as a named parameter with illustrative values printed beside it so the arithmetic can be redone with the real figures substituted. Treat every number above as illustrative of a mechanism and none of it as a statement of the schedule.
How the numbers here were produced. The session is 375 minutes. The illustrative requirement path, the funded level of 3,20,000 and the collateral figures are inputs stated on the page. Minutes short, the largest shortfall and the funding policy comparison are computed from that path directly. The catch probabilities are one minus the probability that every one of k independently drawn snapshots misses a window of m minutes in 375. The monthly figures assume the same behaviour on each of 20 sessions. The collateral figures apply the two conditions stated in the text, being total accepted value against the requirement and cash against the cash component. The repetition figures are expressed in units of a single base charge and are computed for a range of multipliers rather than for any published one.
The position is stated as at September 2026. Margin rules, penalty schedules, haircuts and collateral policies are revised and differ in the detail a member applies. Confirm the current figures with your own clearing member and in the circulars it cites before sizing anything against them, and take advice on your own circumstances.
Ready to go deeper than this article?
Bharath Shiksha is a 90-volume curriculum across 6 stages, from chart reading at ₹14,999 through capital raising, or the full bundle at ₹1,49,999. Margin is taught here as a sizing constraint read before the order rather than as a charge explained after the statement, which is the only point at which it can still be changed.
Take the free diagnostic →