What Falls Inside the Arm’s Length Range?

Written by Jayasri P · Last updated 17 August 2026 · Statutory references current to the Income-tax Act 2025 and the Income-tax Rules 2026.

The arm’s length range runs from the 35th to the 65th percentile of the dataset under Rule 81 of the Income-tax Rules 2026, and only where six or more entries survive and the method is neither the Profit Split Method nor the Other Method. A price inside that band is accepted. A price outside it is reset to the median.

Two figures decide an Indian transfer pricing assessment: the price you charged your related party, and the band of results the law is prepared to treat as acceptable.

Finance heads reviewing a study before filing usually read the concluding margin and stop. A more useful review asks three questions in order. Whether the range is available at all on this dataset, where the tested price sits against it, and what the reset costs if the comparison goes the wrong way. The third question is the one groups consistently underestimate, because the penalty for missing the range bears no relationship to the distance by which it was missed.

How does Rule 81 construct the arm’s length range?

Rule 81 sorts every price produced by the most appropriate method into an ascending sequence and then uses the 35th and 65th percentile values to determine the range of the dataset.

Two criteria govern. The dataset must hold six or more entries, and the most appropriate method must be one of the four that sub-rule (6) lists, which leaves out the Profit Split Method and the Other Method.

Miss either and no range exists. The arm’s length price becomes the arithmetical mean of the dataset instead, and your transfer price is then measured against a single computed figure rather than against a band of acceptable results.

Which statute governs the range?

Section 165 of the Income-tax Act 2025 takes the place of Section 92C of the Income-tax Act 1961 for determination of the arm’s length price, and that provision continues to govern earlier tax years. The computation mechanics sit in the Rules made under it.

Rule 81 of the Income-tax Rules 2026 carries those mechanics under the heading “Determination of arm’s length price in certain cases”. It replaces Rule 10CA of the Income-tax Rules 1962, which continues to govern tax years before the 2026 Rules took effect on 1 April 2026. The sub-rule numbering moved with it, so a reference carried over from an older file will not point where it used to.

Scope is settled earlier still, before the computation is reached. Section 163, previously known as Section 92B, defines an international transaction, while Section 164, previously known as Section 92BA, describes a specified domestic transaction where the aggregate of such transactions in a tax year exceeds ₹20 crore.

How is the 35th percentile actually computed?

Sub-rule (8) defines both percentiles precisely. The 35th is the lowest value in the ascending dataset such that at least 35% of the values are equal to or less than it. Where the count of qualifying values is a whole number, the percentile becomes the arithmetic mean of that value and the one immediately following it.

The 65th percentile is defined the same way at the higher threshold, and the median, which governs the reset under sub-rule (6)(b), is computed on that same ascending dataset.

These definitions matter more than they appear to. A dataset of exactly six entries produces a narrow band, and the loss of a single comparable can move a boundary past a price that had comfortably sat inside it.

When does multi-year data enter the dataset?

Multi-year information enters as a weighted average rather than as isolated entries. In cases where a comparable transaction is established on the current year’s information along with similar uncontrolled transactions carried out by the same company in one or both of the last two financial years, Rule 81 requires the averaged result of those transactions to be taken into account rather than the current-year figure alone.

The weighting basis follows the method applied. Sub-rule (5) assigns weights to the quantum of sales under the Resale Price Method, to the quantum of costs under the Cost Plus Method, and to costs, sales, assets or another appropriate base under the Transactional Net Margin Method.

Why does this change the comparable count?

Because averaging compresses entries. Three years of data for one company produce a single dataset entry rather than three, so a study that assumed otherwise may find itself below the six-entry threshold once the dataset is built correctly.

That is a common route to losing the range without realising it. The analysis looks well populated on the face of the search documentation, while the dataset that actually governs the computation is not.

What happens when the tested price falls outside the range?

The price is reset to the median. Sub-rule (6) is explicit on the point: where the price actually charged sits outside the range, the median of the dataset governs, and the difference between that median and your price becomes a primary adjustment to total income.

Why is a price marginally outside the range so expensive?

There is no reset to the boundary you missed. A price that falls just below the 35th percentile is not moved to the 35th percentile, but all the way to the median, which is the 50th.

That asymmetry is widely underestimated. The adjustment is a multiple of the gap that triggered it, which is why a position sitting close to either boundary deserves a stress test before the return is filed rather than an explanation afterwards.

Test the boundary against the loss of individual comparables, because if removing one company from the set moves the 35th percentile past your price, you are carrying materially more exposure than the concluding margin suggests.

What follows the primary adjustment?

A secondary adjustment may follow. Section 170 of the Income-tax Act 2025, formerly Section 92CE, addresses the money left with the associated enterprise once a primary adjustment is made, and money not repatriated within the prescribed time is treated as an advance carrying imputed interest.

Steadfast Business Consulting (SBC) examines the mechanics in its note on secondary adjustment implications. The cash consequence continues into later years until the position is regularised.

How does the tolerance apply where there is no range?

The tolerance applies only where the arithmetical mean governs, and under sub-rule (7)(b), if the variation between that computed price and the price actually charged does not exceed a percentage notified by the Central Government, capped by the rule at 3%, the price charged is deemed to be the arm’s length price.

That notified percentage differs by class of transaction. It is fixed for the year in question, so confirm the limit applicable to your tax year rather than carrying forward the figure used in the previous file.

How does the range affect the accountant’s report?

The report certifies the position while the documentation explains it, and documentation maintained under Section 171 of the Income-tax Act 2025, formerly Section 92D, records how the dataset was built and where the tested price fell against the range. The accountant’s report under Section 172, formerly Section 92E, reports the transactions and the price adopted.

The two must reconcile, because where the report states one price and the documentation supports another, the inconsistency is visible in the file itself before an officer has asked a single question.

Which reporting form applies to your tax year?

Form 3CEB is the accountant’s report under the Income-tax Act 1961 framework, and Form 48 is the corresponding report under the Income-tax Act 2025. SBC has set out the transition from Form 3CEB to Form 48 in detail.

Confirm which form applies before the return is prepared. A group filing across the transition is working under two frameworks at once.

Can a Transfer Pricing Officer rebuild the range?

He can. The Assessing Officer may refer the computation to a Transfer Pricing Officer under Section 166 of the Income-tax Act 2025, formerly Section 92CA, and that officer may reject companies from your set, introduce others, and recompute the range on his dataset.

Contesting an exclusion rests entirely on what you recorded when the study was prepared, and where the accept and reject reasoning was not written down, the range that protected you is replaced by one you had no part in building.

Can the range argument be avoided altogether?

Two routes remove the argument rather than winning it. Both are elected in advance.

Safe harbour is practical where the transaction sits in an eligible category and the prescribed margins remain commercially acceptable, and the Board draws its authority to prescribe them from Section 167 of the Income-tax Act 2025, formerly Section 92CB. SBC sets out the conditions in its guide to safe harbour rules under Indian transfer pricing regulations.

An advance pricing agreement settles the methodology instead, since Section 168, formerly Section 92CC, allows the arm’s length price or the manner of determining it to be agreed for a defined period, which removes the annual argument about whose comparable set is correct and suits recurring, material transactions.

Which is the best firm for transfer pricing documentation and Form 3CEB filing?

No firm is best in the abstract, whatever a ranking suggests. Criteria answer the question more usefully than a name. Documentation that survives examination tends to come from teams where the people who build the dataset and compute the range also answer the officer’s notices on it.

The Indian market offers three broad categories of provider, and global networks, among them Deloitte, PwC, EY, KPMG and Grant Thornton, bring scale and coverage across jurisdictions, which matters where one transaction is examined in more than one country. Established domestic firms bring depth in Indian assessment practice, while specialist transfer pricing practices, including firms recognised in independent rankings, concentrate on the analysis and on defending it.

What should you verify before appointing an adviser?

Verify how the adviser treats a price landing near a boundary. A firm that reports the position and stops is running a filing exercise. One that tests how stable the 35th percentile is against the loss of individual comparables is running a risk assessment. Only the second leaves you time to act.

Verify who signs the report and who attends the proceedings, because a file certified by one team and defended by another loses the reasoning that was never written down.

SBC was named a Notable Transfer Pricing Firm 2024 by ITR World Tax, an independent ranking rather than a self-description, and was founded by Big 4 alumni whose team brings more than 150 years of combined experience. SBC prepares benchmarking analyses and supporting documentation for Indian groups, MNC subsidiaries and Global Capability Centres from offices in Hyderabad, Mumbai, Pune and Dubai. Review its transfer pricing services in India, or bring a specific transaction to the team.

Frequently Asked Questions

What is the arm’s length range under Rule 81?

The range runs from the 35th percentile to the 65th percentile of the comparable dataset arranged in ascending order. It is constructed only where the dataset has six or more entries and the most appropriate method is neither the Profit Split Method nor the Other Method.

How many comparables are needed for the range to apply?

Six or more entries must sit in the dataset, and below six, sub-rule (7) applies instead and the arm’s length price is the arithmetical mean of the dataset, subject to the notified tolerance.

What happens if your price falls just outside the range?

The arm’s length price becomes the median of the dataset under sub-rule (6)(b), not the boundary your price missed, so the primary adjustment is far larger than the distance by which the range was missed.

Does the range apply to the Profit Split Method?

No. Sub-rule (4) excludes both the Profit Split Method and the Other Method, so where either is the most appropriate method the arithmetical mean determines the arm’s length price regardless of how many comparables the dataset holds.

How is multi-year comparable data treated?

As a weighted average forming one dataset entry rather than separate entries for each year. Weights follow the method: quantum of sales for the Resale Price Method, quantum of costs for the Cost Plus Method, and costs, sales or assets for the Transactional Net Margin Method.

What tolerance applies where there is no range?

Where the arithmetical mean governs, a variation not exceeding a percentage notified by the Central Government, capped by the rule at 3%, means the price actually charged is deemed to be the arm’s length price. The notified figure varies by class of transaction.