Guide · Execution and market structure · Updated 2026-09-23

The spread prices a token order. Impact cost prices yours, and it is the number the index rules actually use.

Trading screens show the spread. Few show the number that decides what a trade of real size costs, although it was defined when the exchange's benchmark index was designed, is still used to decide which companies may enter that index, and is published for the index itself. This guide works the definition from the exchange's own examples, runs it across sizes on both sides of a stated book, checks the index rule as it stands in September 2026, and measures how badly the most quoted liquidity statistic ranks depth. The guide to market depth reads the ladder; this one is about the number built from it.

A mark-up against the midpoint, not a gap between two quotes

Impact cost is the percentage mark-up suffered while buying or selling the desired quantity of a stock, measured against its ideal price: the best buy price plus the best sell price, divided by two. For a purchase, take the quantity-weighted average of the offers the order consumes, subtract the ideal price and divide by it; for a sale, the same arithmetic runs down the bids. The source is the Impact Cost section of the NSE Indices methodology document for equity indices, June 2026 edition, repeated on the exchange's own impact cost page.

Four qualifications follow, and they are the substance. Impact cost is computed separately for buy and sell, varies with order size, changes with the orders outstanding, and is replaced by a penal figure where a stock is not sufficiently liquid. An order larger than the book, which has no average price at all, is the plainest such case.

The exchange's worked example has a best bid of 3.50 and a best offer of 4.00, so the ideal price is 3.75. A sale of 4,000 shares takes 1,000 at 3.50, 1,000 at 3.40 and 2,000 at 3.40, an average of 3.425.

Unrounded, the sale costs 8.67 per cent. The exchange prints 8.53, because its table rounds the average to 3.43 before dividing, and the sentence stating the result calls the order a purchase although every line of the table is a sale. Neither slip changes the method. Both are reasons to recompute a published example rather than copy it.

The spread is the first rupee, and a round trip pays it twice

Most of that 8.67 per cent is not the order's size at all: 6.67 points are half the spread, paid before a single share trades by every sale however small, and only 2.00 come from walking down to 3.40.

That holds for any book. An order that fits inside the quantity at the best price has an impact cost equal to half the spread, as a fraction of the ideal price, whatever its size within that level. Beyond it, each share fills no better than the one before, so the average can only hold or worsen. The quoted spread is the limit of impact cost as the order shrinks to nothing, which makes it the one figure on the screen that says nothing about your size. The guide to the bid-ask spread covers what sets it; this is what it leaves out.

A round trip pays both halves and both walks. In the exchange's book the spread is 13.33 per cent of the ideal price; buying 3,000 shares costs 7.11 per cent and selling them 8.44, a round trip of 15.56. In the methodology's second example, with a best bid of 98 and a best offer of 99, buying 1,500 shares averages 99.33, an impact cost of 0.846 per cent, of which 0.508 is half the spread. The reference is the midpoint rather than the last trade because the last print can sit anywhere between bid and offer, while the midpoint charges buyer and seller symmetrically, which is what lets the two sides of one book be compared.

One book, both sides, every size

The exchange's examples are small and extreme, with spreads of 13 per cent and 1 per cent of the price. To see the shape at the sizes traders use, take a stated book: twelve levels a side around 640, a spread of 20 paise, and a bid side written deliberately thinner and more gapped than the offers. It is illustrative, and set out in full so every figure below can be redone by hand.

At ₹1 lakh every column equals half the 0.031 per cent spread. By ₹50 lakh the purchase costs 0.120 per cent and the sale 0.156, and half the spread is down to 13 per cent of the purchase's cost. The sale is 30 per cent dearer at the same size in the same stock, which is why the definition refuses to give one number. At ₹1 crore the sale cannot complete, because the visible bids hold ₹93.8 lakh: no average price exists, the plainest case of a stock not being sufficiently liquid, for which the methodology applies a penal figure it does not publish.

Why the bill outruns the order

The rupee bill is the percentage applied to the order, and the percentage itself rises with size, so the bill rises faster than the order. On the illustrative book a ₹25 lakh purchase pays ₹1,509 in impact and a ₹50 lakh purchase ₹6,013: twice the order, 4.0 times the bill. The exchange's own FAQ makes the point with round numbers: 1,000 shares filled 2 per cent above an ideal price of 100 cost ₹2,000, and 2,000 shares filled 4 per cent above it cost ₹8,000.

The percentage is also an average, and the next rupee of size does not pay the average. The last shares of the ₹50 lakh purchase filled at 641.50, 0.234 per cent above the ideal price, 1.9 times the 0.120 per cent average. On a book whose depth is roughly even across ticks the marginal concession grows about twice as fast as the average, so adding to a position at size is dearer than its average cost suggests.

How steeply the percentage rises depends on the book's shape: depth that thins away from the touch makes it rise faster than size, depth that thickens more slowly, and the guide to iceberg orders and market impact fits the exponent on three shapes. The rupee bill outruns the order on every one of them.

The index rule tests a basket of ₹10 crore, split by weight

The Nifty 50 turns the number into a gate. Under the June 2026 methodology a stock qualifies only if it sits in the Nifty 100, has derivatives available on the exchange, traded on every day of the previous six months, and showed an average impact cost of 0.50 per cent or less in 90 per cent of observations over those months, for a portfolio of ₹10 crore. A newcomer also needs 1.5 times the free-float market capitalisation of the smallest constituent. The factsheet dated 31 August 2026 states the same basket as 100 million rupees.

The basket was ₹2 crore until the index provider's press release of 22 August 2017 raised it to ₹10 crore, from the September 2017 rebalancing. The exchange's own FAQs about Indices page, as fetched on 23 September 2026, still describes Nifty 50 trades of two crore rupees, and that page is where the old figure keeps entering explainers.

The second correction changes what the test measures. The methodology measures each security's transaction "in proportion to its index weight": the ₹10 crore is the whole index bought as a basket, and each stock is tested with its slice. The largest constituent, 9.85 per cent on the August 2026 factsheet, is tested with about ₹98.5 lakh. A stock with half a per cent is tested with ₹5 lakh, and before September 2017 would have been tested with ₹1 lakh. It is not a ₹10 crore order in each stock, which is how the rule is often paraphrased.

Nor does the screen stop at fifty names. The Nifty 500's eligible universe, from which the Nifty 100 and the mid and small cap indices are drawn, excludes any company whose average impact cost exceeded 1 per cent over the previous six months, and so do the Nifty Microcap 250 and Nifty Smallcap 500. The test was absent from the November 2019 broad market methodology, was present in the February 2022 edition, and carries no stated basket size.

The exchange also publishes the index's own figure: 0.02 per cent for a ₹50 lakh portfolio for March 2026, on its Nifty 50 page. The FAQ page that carries the old basket still puts such a trade at about 0.1 per cent, with the index at 1,000. The rest of the index rule is covered in the guide to the Nifty 50.

The derivatives test asks the same question backwards

Stock derivatives carry their own depth test, the index test turned inside out. Under SEBI circular SEBI/HO/MRD/MRD-PoD-2/P/CIR/2024/116 of 30 August 2024, a stock needs a median quarter sigma order size of at least ₹75 lakh over six months, up from ₹25 lakh. That is the rupee order needed to move the price by a quarter of one daily standard deviation, the exponentially weighted figure the exchange uses for margins. The exchange's derivatives selection criteria page gives the procedure: four order book snapshots a day, each side computed at every snapshot, the median of each side over six months, and the average of the two medians.

The index test fixes a size and caps the cost; the derivatives test fixes a cost and floors the size. The exchange publishes the second monthly, security by security. Its file for 16 February to 14 August 2026 covers 383 securities, with a median of ₹3.53 crore, a thinnest tenth below ₹82.8 lakh and a smallest of ₹27.8 lakh.

A ₹5 lakh slice is 1.4 per cent of that median and under a fifth of the smallest. At a snapshot where an order is smaller than the quarter sigma size, no share of it fills more than a quarter of a daily standard deviation from the midpoint. For a stock that already has derivatives, as every Nifty 50 candidate must, the 0.50 per cent screen therefore looks light, and the demanding liquidity test is the one inside derivatives eligibility. That is an inference from scale, not a count: the per stock series behind the index screen is not published, so how often it binds cannot be measured from outside.

Volume is a flow, depth is an inventory

Share volume counts what changed hands over a session; depth is what rests in the book at an instant. They connect only through how fast the book refills, and the published file shows how far apart they are. Across the 383 securities, a stock's quarter sigma order size is a median 2.08 per cent of its average day's turnover, with the middle eighty per cent between 0.98 and 3.17 per cent: the order that moves the median stock a quarter of a day's standard deviation is about one fiftieth of what trades in it by the close. The security bhavcopy for the same 122 sessions gives five daily statistics for each security, and each can be ranked against the published figure.

Shares traded a day, the default sort on most screens, ranks depth at 0.41. Of the 50 securities with the most shares traded, 19 are among the 50 deepest and 12 sit in the shallower half. A security priced near ₹34 ranked 19th of 383 by shares traded and 359th by depth. The whole difference is price: shares traded times the average price is turnover, so a share count treats a ₹30 share and a ₹3,000 share as one unit. Turnover, the same activity in rupees, reaches 0.90, puts 39 of its top 50 among the 50 deepest, and moves 6.5 per cent of securities more than 100 places, against 38.9 per cent for share volume.

The best of the five needs no book at all. The average absolute daily move divided by rupee turnover, a form of the illiquidity ratio long used in research, ranks the published measure at -0.94, and dividing each stock's depth by its own volatility leaves the ordering intact. Two controls stand behind these figures: the sessions match the daily index files one for one, and the file's own average daily deliverable value, recomputed from the bhavcopy, agrees at a rank correlation of 0.997.

A busy fortnight is not a deep book

Turnover ranks depth well on average and badly in one identifiable case, when it arrives in a few sessions. Hold turnover fixed and the depth it leaves unexplained correlates at -0.52 with the share of six months' turnover that fell on the three busiest sessions. Grouped into fifths, depth per rupee of turnover falls steadily as turnover concentrates.

The extreme case is a security with 55.0 per cent of its six months' turnover on three sessions, its busiest day 36 times its average. It ranked 218th by turnover and 381st of 383 by depth. Large single prints do not flag the same thing: at equal turnover, average trade size is weakly positively related to depth, +0.20. Newly traded securities show the pattern most plainly. The 25 that first appear in the file after 16 February 2025 under every symbol they have used, which is how new listings and demerged businesses enter it, have a median depth of 0.84 per cent of a day's turnover against 2.15 per cent for the rest, and make up 9 of the 20 widest gaps between turnover rank and depth rank.

Size the position by the exit

An entry can be patient: a limit order can wait and let the book come to it. An exit often cannot. A stop, a margin call or a piece of news asks for immediacy when resting bids are thinnest, because the traders who were posting them are selling too. The binding constraint on size is the sale, at size, into a thinned book.

On the illustrative book the largest purchase within a 0.10 per cent budget is ₹42.2 lakh, the largest sale ₹33.1 lakh, and the largest sale once the bids halve ₹16.6 lakh, under two fifths of what the entry allowed. A position sized to what the book let you buy is larger than the book will let you sell in a hurry.

The published file gives that arithmetic real scale, and its figures are medians, so on roughly half the snapshots each side of the book was thinner than shown.

These are the top of the market by capitalisation and traded value. The other 1,735 securities that traded on at least nine in ten of the same sessions had a median turnover of ₹5.42 crore a day, against ₹147.9 crore inside the list, where depth tracks turnover almost one for one. Working an order over time changes the arithmetic at the price of time and exposure, which the guides to VWAP and TWAP execution and to slippage take up.

What the number cannot tell you

Impact cost is computed on a snapshot of intentions that can be withdrawn before an order arrives. Disclosed quantity orders show only part of their size, so the visible book is a floor on what rests, not a ceiling. A book refills between trades, so an order worked across a session pays less than one sweep of the same size and pays in timing risk instead. And the figure is the immediate cost of crossing, not the lasting move a large worked order leaves in the price.

The published figures have limits of their own. The index number is for one month and one basket size, the per stock series behind the index screens is not released, and the quarter sigma file is a six month median. Neither the ₹10 crore basket nor the ₹75 lakh floor says what a particular trade will cost on a particular morning.

What the number is for

Used properly, impact cost replaces one misleading habit with three useful ones. Compare securities by what rests in the book, or by rupee turnover and price response when the book cannot be seen, never by shares traded. Price a trade by its size on its own side of the book, both sides for a round trip, never by the quoted spread. Set position size by what can be sold at once into a thinned book, not by what could be bought patiently. Treating cost as a property of your order in a particular book, rather than a line on a contract note, is the kind of execution judgement the curriculum teaches as method, and the full cost stack of an Indian trade sits on top of it.