Press / to search
Rule 81Published

Rule 81: Arm’s-length range, weighted averages and median

When the most appropriate transfer-pricing method produces more than one price, Rule 81 sets the dataset and calculation framework. The applicable method, number of entries, year of data and reliability of comparable transactions determine whether the range or arithmetic-mean approach applies.

Official notified Income-tax Rules, 2026

Need advice on Rule 81?

This article is for general information. Read the official Rules and applicable Act amendments, including the Taxation and Other Laws (Amendment) Act, 2026. If you need professional advice for your situation, contact Bizeneed.

Contact us
On this page

Bare law

Rule 81 - Determination of arm's length price in certain cases

(1) Where in respect of an international transaction or a specified domestic transaction, the application of the most appropriate method referred to in section 165(3)(b) results in determination of more than one price, the arm‘s length price in respect of such transactions shall be computed in accordance with the provisions of this rule.

(2) A dataset shall be constructed by placing the prices referred to in sub-rule (1) in an ascending order and the arm's length price shall be determined on the basis of the dataset so constructed.

(3) For the purposes of sub-rule (2), where the most appropriate method is the resale price method, cost plus method, or transactional net margin method, and where the comparable uncontrolled transaction undertaken by an enterprise, not being the enterprise undertaking such transactions referred to in sub-rule (1), has been identified using data relating to— (a) the current year, and the said enterprise has undertaken the same or similar comparable uncontrolled transaction, in either or both of the two financial years immediately preceding the current year, then, — (i) the price in respect of such transaction in such year (or years) shall be determined by applying the most appropriate method in the similar manner as applied in the current year; and (ii) the weighted average of such prices, as computed under sub-rule (5), shall be included in the dataset under sub-rule (2) instead of price referred to in sub-rule (1); (b) the financial year immediately preceding the current year (if the data relating to current tax year is not available at the time of furnishing return of income for that year), and the said enterprise has undertaken the same or similar comparable uncontrolled transaction in the financial year immediately preceding the two financial years, then— (i) the price in respect of such transaction shall be determined by applying the most appropriate method in the similar manner as it was applied in the financial year immediately preceding the current year; and (ii) the weighted average of such prices, as computed under sub-rule (5), shall be included in the dataset under sub-rule (2) instead of price referred to in sub-rule (1).

(4) Where the use of data relating to the current year, in terms of rule 79(5), establishes that— (a) the enterprise has not undertaken the same or similar uncontrolled transaction during the current year; or (b) the uncontrolled transaction undertaken during the current year is not a comparable uncontrolled transaction, then, irrespective of anything else, neither the price nor the weighted average of prices of the comparable uncontrolled transactions shall be included in the dataset.

(5) Where an enterprise has undertaken comparable uncontrolled transactions in more than one financial year, the weighted average of the prices of such transactions for the purposes of sub-rules (2), (3) and (4) shall be computed based on the method used for determination of prices as specified in column B of the following Table and by assigning weights to the factors specified in column C thereof: —

Sl. No. (A)Method (B)Factors to which weight is assigned (C)
1.Resale price method.Quantum of sales considered for arriving at the respective prices.
2.Cost plus method.Quantum of costs considered for arriving at the respective prices.
3.Transactional net margin method.Quantum of costs incurred, sales effected, assets employed or to be employed, or any other base considered for arriving at the respective prices.

(6) Where the most appropriate method is comparable uncontrolled price method or resale price method or cost plus method or transactional net margin method and the dataset constructed in accordance with sub-rule (2) consists of six or more entries, an arm's length range beginning from the 35th percentile of the dataset and ending on the 65th percentile of the dataset shall be constructed and, if the price at which the international transaction or the specified domestic transaction has actually been undertaken is — (a) within such arm‘s length range, such price shall be deemed to be the arm's length price; (b) outside such arm's length range, the median of the dataset shall be used to compute the arm‘s length price.

(7) Where the provisions of sub-rule (6) are not applicable, the arm‘s length price shall be — (a) the arithmetical mean of all the values included in the dataset; or (b) the price at which such transaction has actually been undertaken, if the variation between the arm‘s length price so determined and the price at which the international transaction or the specified domestic transaction has actually been undertaken does not exceed such percentage, not exceeding 3% of the latter, as may be notified in this behalf by the Central Government.

(8) For the purposes of this rule, — (a) ―median‖ of the dataset, having values arranged in an ascending order, shall be — (i) the lowest value in the dataset where at least 50% of the values are less than or equal to it; or (ii) the arithmetic mean of such lowest value and the value immediately succeeding it in the dataset, if the number of all values that are equal to or less that the aforesaid value is a whole number; (b) (i) ―35th percentile‖ of a dataset, having values arranged in an ascending order, shall be — (A) the lowest value in the dataset such that at least 35% of the values included in the dataset are equal to or less than such value; or (B) the arithmetic mean of such lowest value and the value immediately succeeding it in the dataset, if the number of all values that are equal to or less that the aforesaid value is a whole number; and (ii) ―65th percentile‖ of a dataset, having values arranged in an ascending order, shall be — (A) the lowest value in the dataset such that at least 65% of the values included in the dataset are equal to or less than such value; or (B) the arithmetic mean of such lowest value and the value immediately succeeding it in the dataset, if the number of all values that are equal to or less that the aforesaid value is a whole number.

Illustration 1 Enterprise X has undertaken controlled transaction with its associated enterprises during the current year. After taking into account the facts and circumstances, transactional net margin method has been selected as the most appropriate method, enterprise X has been selected as the tested party, and the ratio of operating profit (OP) to operating expense (OE) has been selected as the profit level indicator (PLI). The data of the comparable uncontrolled transactions is available for the current year under consideration at the time of furnishing return of income by the assessee and based on the same, seven non-associated enterprises have been identified to have undertaken the comparable uncontrolled transactions in the current year. All the identified comparable enterprises have also undertaken comparable uncontrolled transactions in a period of two years preceding the current year. The weighted average PLI calculation for each non-associated enterprise shall be as follows:

Sl. No.Name of non-associated enterpriseYear 1Year 2Year 3 [Current Year]Aggregation of OE and OPWeighted Average PLI
1.AOE = 100; OP = 12OE = 150; OP = 10OE = 225; OP = 35Total OE = 475; Total OP = 57OP/OE = 12%
2.BOE = 80; OP = 10OE = 125; OP = 5OE = 100; OP = 10Total OE = 305; Total OP = 25OP/OE = 8.2%
3.COC=E = 250; OP = 22OE = 230; OP = 26OE = 250; OP = 18Total OE = 730; Total OP = 66OP/OE = 9%
4.DOE = 180; OP = (-)9OE = 220; OP = 22OE = 150; OP = 20Total OE = 550; Total OP = 33OP/OE = 6%
5.EOE = 140; OP = 21OE = 100; OP = (-)8OE = 125; OP = (-)5Total OE = 365; Total OP = 8OP/OE = 2.2%
6.FOE = 160; OP = 21OE = 120; OP = 14OE = 140; OP = 15Total OE = 420; Total OP = 50OP/OE = 11.9%
7.GOE = 150; OP = 21OE = 130; OP = 12OE = 155; OP = 13Total OE = 435; Total OP = 46OP/OE = 10.57%

From the above, the dataset will be constructed by arranging the weighted average values of PLI in an ascending order as follows:

Sl. No.1234567
Values2.2%6%8.2%9%10.57%11.9%12%

For deriving the arm's length range, the data place of 35th and 65th percentile shall be computed in the following manner: Total number of data points in dataset × (35/100) Total number of data points in dataset × (65/100) Thus, the data place of the 35th percentile = 7 × 0.35 = 2.45. Since this is not a whole number, the next higher data place, that is, the value at the 3rd place would have at least 35% of the values below it. The 35th percentile, is therefore, value at the 3rd place, that is, 8.2%. The data place of the 65th percentile is = 7 × 0.65 = 4.55. Since this is not a whole number, the next higher data place, that is, the value at the5th place would have at least 65% of the values below it. The 65th percentile is therefore value at 5th place, that is, 10.57%. The arm's length range will be beginning at 8.2% and ending at 10.57%. Therefore, if the PLI in case of the enterprise X that has undertaken the international transaction or the specified domestic transaction is equal to or more than 8.2% and less than or equal to 10.57%, the price at which such international transaction or the specified domestic transaction has actually been undertaken shall be deemed to be the arm's length price. No adjustment shall be required. However, if the PLI in case of the enterprise X is lower than the 35th percentile of the arm's length range, that is, 8.2%, then for the purpose of determining the arm's length price, the median of the dataset shall be used. The median will be first determined in the following manner: The data place of median is calculated by first computing the total number of data point in the dataset × (50/100). In this case it is 7 × 0.5 = 3.5. Since this is not a whole number, the next higher data place, that is, the value at the 4th place in the dataset would have at least 50% of the values below it (median). The median in this case is 9%. Therefore, the PLI at arm‘s length shall be 9% and the controlled transaction shall be adjusted so as to yield this PLI for the controlled enterprise X.

Illustration 2 In case of the Enterprise X referred to in Illustration 1, the data of the current year is available only in respect of non-associated enterprises A, C, E, F and G at the time of furnishing the return of income by the assessee. In respect of non-associated enterprises B and D, the data of the financial year preceding the tax year is available at the time of furnishing the return of income and the same has been used to identify comparable uncontrolled transactions undertaken by enterprises B and D. Further, if these seven non-associated enterprises have also undertaken comparable uncontrolled transactions in either or both of the two financial years immediately preceding the tax year then such data has been included in the dataset. The weighted average PLI calculation for each non-associated enterprise shall be as follows:

Sl. No.Name of non-associated enterpriseYear 1Year 2Year 3 [Tax Year]Aggregation of OE and OPWeighted Average PLI
1AOE = 100; OP = 12OE = 150; OP = 10OE = 225; OP = 35Total OE = 475; Total OP = 57OP/OE = 12%
2BOE = 80; OP = 10OE = 125; OP = 5Data not availableTotal OE = 205; Total OP = 15OP/OE = 7.31%
3COE = 250; OP = 22OE = 230; OP = 26OE = 250; OP = 18Total OE = 730; Total OP = 66OP/OE = 9%
4DData not availableOE = 220; OP = 22Data not availableTotal OE = 220; Total OP = 22OP/OE = 10%
5EData not availableData not availableOE = 125; OP = (-)5Total OE = 125; Total OP = (-)5OP/OE = (-)4%
6FOE = 160; OP = 21OE = 120; OP = 14OE = 140; OP = 15Total OE = 420; Total OP = 50OP/OE = 11.9%
7GOE = 150; OP = 21OE = 130; OP = 12OE = 155; OE = 13Total OE = 435; Total OP = 46OP/OE = 10.57%

From the above, the dataset will be constructed by arranging the weighted average values of PLI in an ascending order as follows:

Sl. No.1234567
Values(-)4%7.31%9%10%10.57%11.9%12%

If during the course of assessment proceedings, the data of the tax year becomes available and the use of such data indicates that non-associated enterprise B cannot be considered as a comparable under rule 79(3)(ii) and (iii), then the data in respect of enterprise B shall not be included in the dataset. The data for tax year in respect of enterprise D shall be included in the dataset. Further, if the data available at this stage identifies a new comparable uncontrolled transaction undertaken by enterprise H, then, it shall be included in the revised dataset. The weighted average PLI calculation for each non-associated enterprise shall be as below:

Sl. No.Name of non-associated enterpriseYear 1Year 2Year 3 [tax Year]Aggregation of OE and OPWeighted Average PLI
1AOE = 100; OP = 12OE = 150; OP = 10OE = 225; OP = 35Total OE = 475; Total OP = 57OP/OE = 12%
2COE = 250; OP = 22OE = 230; OP = 26OE = 250; OP = 18Total OE = 730; Total OP = 66OP/OE = 9%
3DData not availableOE = 220; OP = 22OE = 150; OP = 20Total OE = 370; Total OP = 42OP/OE = 11.35%
4EData not availableData not availableOC = 125; OP = (-)5Total OC = 125; Total OP = (-)5OP/OC = (-)4%
5FOC = 160; OP = 21OC = 120; OP = 14OC = 140; OP = 15Total OC = 420; Total OP = 50OP/OC = 11.9%
6GOC = 150; OP = 21OC = 130; OP = 12OC = 155; OP = 13Total OC = 435; Total OP = 46OP/OC = 10.57%
7HOC = 150; OP = 12Data not availableOC = 80; OP = 10Total OC = 230; Total OP = 22OP/OC = 9.56%

From the above, the dataset will be constructed by arranging the weighted average values of PLI in an ascending order as follows:

Sl. No.1234567
Values(-)4%9%9.56%10.57%11.35%11.9%12%

The calculation of the arm‘s length range and the arm‘s length price shall then be performed in the same manner as that described in Illustration 1.

Illustration 3 In a given case, comparable uncontrolled price method has been selected as the most appropriate method and 20 comparable uncontrolled transactions have been identified. The dataset of 20 prices, arranged in ascending order, is as under:

Sl. No.Profits (in ₹ Thousands)
142.00
243.00
344.00
444.50
545.00
645.25
747.00
848.00
948.15
1048.35
1148.45
1248.48
1348.50
1449.00
1549.10
1649.35
1749.50
1849.75
1950.00
2050.15

Applying the formula given in the Illustration 1, the data place of the 35th and 65th percentile is determined as follows: 35th percentile place = 20 × (35/100) = 7th. 65th percentile place = 20 × (65/100) = 13th. Since the 35th percentile place is a whole number, it shall be the average of the prices at the 7th and next higher, that is, 8th place. This is (47000+48000)/2 = ₹ 47500. Similarly, the 65th percentile will be average of 13th and 14th place prices. This is (48500+49000)/2 = ₹ 48750. The median of the range (the 50th percentile place) = 20 × (50/100) = 10th Since the 50th percentile place is a whole number, it shall be the average of the prices at the 10th and next higher, that is, 11th place. This is (48350+48450)/2 = ₹ 48400. Thus, the arm's length range in this case shall be from ₹ 47500 to ₹ 48750. Consequently, any controlled transaction whose price is equal to or more than ₹ 47500 but less than or equal to ₹ 48750 shall be considered to be within the arm's length range. If such price is not within the arm‘s length range, then, for the purpose of determining the arm's length price, the median of the dataset shall be used.

Rule breakdown

Start with an ordered and supportable dataset

Arrange the relevant prices in ascending order. First establish the method and comparability under the preceding rules; a larger dataset does not cure unsuitable comparables.

For resale price, cost plus and TNMM, the rule specifies when comparable transactions from the preceding financial years enter through weighted averages. It also addresses the limited case in which current-year data is unavailable when the return is furnished.

Compute weighted averages using the correct base

Weight resale-price results by sales and cost-plus results by costs. TNMM uses the relevant base, such as costs, sales or assets, used to arrive at the prices.

An unweighted average of annual percentages can give a different result. In official Illustration 1, enterprise A has aggregate operating profit of 57 and operating expenses of 475, giving 12%; the calculation follows the underlying amounts.

Reassess comparables when current-year data becomes available

If current-year data establishes that the enterprise did not undertake the same or a similar uncontrolled transaction, or that the transaction is not comparable, neither its price nor its weighted average belongs in the dataset.

Official Illustration 2 demonstrates removal of an unsuitable comparable, updating another enterprise’s current-year data and addition of a newly identified comparable. Keep the reasons and the revised calculation together.

When the 35th–65th percentile range applies

For CUP, resale price, cost plus or TNMM, a dataset containing at least six entries triggers the specified range from the 35th to the 65th percentile.

A transaction price within the range is treated as the arm’s-length price. If it is outside the range, use the dataset’s median to compute the arm’s-length price; do not simply adjust to the nearer endpoint.

Use the percentile convention shown in the statutory illustrations

Sort the dataset and calculate the data place by multiplying its size by 35%, 65% or 50%. The official illustrations use the next higher place where that result is not a whole number, and average the value at that place and the following place where it is a whole number.

For the seven-entry dataset in Illustration 1, the endpoints are 8.2% and 10.57%, and the median is 9%. These figures illustrate that dataset; they are not general acceptable margins.

Worked price example: the official twenty-entry dataset

Official Illustration 3 gives a 35th-percentile place of 7 and a 65th-percentile place of 13. Averaging entries 7 and 8 gives ₹47,500, and averaging entries 13 and 14 gives ₹48,750.

The median averages entries 10 and 11: (₹48,350 + ₹48,450) ÷ 2 = ₹48,400. A controlled price from ₹47,500 to ₹48,750 is within this example’s range; otherwise the median is used for the arm’s-length calculation.

Arithmetic mean and notified tolerance in other cases

Where the range provision does not apply, use the arithmetic mean of the dataset, subject to the actual-price treatment if the variation satisfies the percentage notified by the Central Government.

The rule caps that notified percentage at 3% of the actual transaction price. It does not itself grant a universal 3% deduction or tolerance. Check the notification applicable to the transaction and period.

Practical review before finalising the result

Retain the comparable search, functional analysis, data years, weighted-average workings, sorted dataset and exact percentile calculations. Record why excluded enterprises fail comparability and how later data changes the result.

Check the original notified illustrations alongside the reproduced tables where printed labels are inconsistent. Apply the rule to the actual facts and obtain advice before using an illustrative margin as a filing position.

Key takeaways

At least six entries for the specified range

The 35th–65th range applies to the four named methods and the required dataset.

Outside the range: median

The rule uses the median, not the nearest endpoint.

Weights follow the method

Sales, costs or the relevant TNMM base determine weighting.

No automatic 3% allowance

The alternative calculation depends on the applicable notified percentage.

Practical checklist

  • Document the selected method and comparable dataset.
  • Use the correct weighting base and update current-year evidence.
  • Keep sorted data and percentile workings reproducible.
  • Check the applicable tolerance notification where sub-rule (7) applies.

Avoid

  • Do not use the nearer endpoint when the rule requires the median.
  • Do not treat the statutory illustrations as universal safe-harbour margins.
  • Do not assume a flat 3% reduction from the arm’s-length result.

FAQs

Does every dataset with six entries qualify for the range?

Sub-rule (6) also requires one of the specified methods: CUP, resale price, cost plus or TNMM. The dataset must be constructed under the rule.

If the price is outside the range, can I use the nearer endpoint?

The rule requires the median of the dataset to compute the arm’s-length price.

Can annual margins simply be averaged without weights?

Apply the weights in sub-rule (5). A simple average of annual percentages may not equal the required weighted result.

Is 3% always available as a tolerance?

No. The statutory wording refers to the percentage notified by the Central Government, which cannot exceed 3%, and the specified conditions must be met.

Are the 8.2% and 10.57% figures standard acceptable margins?

No. They come from the seven-entry dataset in official Illustration 1 and are not prescribed general margins.

Glossary

Weighted average
An average using the method-specific sales, cost or other prescribed base as weights.
Arm’s-length range
The 35th–65th percentile range constructed in the circumstances specified in sub-rule (6).
Median
The 50th-percentile measure determined under the rule and its illustrations.
PLI
Profit level indicator; the illustrations use operating profit relative to the stated operating expense or cost base.

Official sources