टीपी भारत
TP Bharat · Data. Insight. Comparables.Explore the databaseAccessibility

Benchmarking under Indian Transfer Pricing law

The legal framework, the comparability standard, multiple-year data, the range and the tolerance, and how each step of a TP Bharat search relates to them. Written for practitioners.

Reviewed 3 October 2026. This page explains how we apply the law. It is not legal advice; check the Act, the Rules and the notifications for the year in question.

Legal framework

The Income-tax Act, 2025 replaced the 1961 Act from 1 April 2026. Transfer Pricing is now in sections 161 to 173, which carry over sections 92 to 92F. The Income-tax Rules, 2026 replace the 1962 Rules, including Rules 10A to 10E. Financial year 2025-26 and earlier years stay under the old Act and Rules; tax year 2026-27 onwards falls under the new ones.

Provision1961 Act2025 Act
Arm’s length principle; no reduction of income92161
Associated enterprise92A162
International transaction92B163
Specified domestic transaction92BA164
Computation of arm’s length price92C165
Reference to the Transfer Pricing Officer92CA166
Safe harbour92CB167
Advance pricing agreement92CC168
Secondary adjustment92CE170
Documentation92D171
Accountant’s report92E172
Definitions92F173

The forms were renumbered as well:

FormOldNew (from tax year 2026-27)
Accountant’s report3CEB48
Safe harbour option3CEFA / 3CEFB49
Master file3CEAA56
Master file intimation3CEAB57
Country-by-country report: intimation / report / other cases3CEAC / 3CEAD / 3CEAE58 / 59 / 60
Multi-year arm’s length price: option / accountant’s certificatenot applicable46 / 47

When the rules apply

Income from an international transaction between associated enterprises is computed having regard to the arm’s length price. The rules also cover specified domestic transactions above ₹20 crore in aggregate, and certain transactions with unrelated parties that are in substance determined between associated enterprises.

The rules work in one direction only. An adjustment cannot reduce the income chargeable in India or increase a loss. This is why a tested margin above the range is left alone.

Methods and profit level indicators

Six methods are available: comparable uncontrolled price (CUP), resale price, cost plus, profit split, the transactional net margin method (TNMM) and any other method that fits. The most appropriate method is chosen on the facts, considering the nature of the transaction, the functions, assets and risks of each party, the availability and reliability of data, and how reliably differences can be adjusted for.

Company databases are used mostly for TNMM, which compares the tested party’s net margin with that of independent companies, measured by a profit level indicator (PLI):

  • Operating profit / total operating cost for service providers and contract manufacturers paid on cost.
  • Operating profit / sales for distributors.
  • Berry ratio (gross profit / operating expenses) for low-risk distributors and agents whose value lies in their operating expenses.
  • Gross profit / sales supports the resale price method where gross margins are comparable.

The denominator should be the base the tested party is actually remunerated on, and it should not be affected by the controlled transaction itself. Pass-through costs that earn no mark-up can be excluded through the value-added expenses PLI.

Tested party

The tested party is normally the less complex party, the one whose functions can be benchmarked reliably and for which data is available. For most Indian captives, contract manufacturers and distributors, that is the Indian entity, and its own audited figures are used. Using a foreign associated enterprise as the tested party needs reliable data for it and comparables from its market.

Comparability

Comparability is judged on four factors: the characteristics of the property or services; the functions performed, taking into account assets used and risks assumed; the contractual terms; and the conditions in the relevant markets, such as geography, size of market, competition and level of the market. An uncontrolled company is comparable if none of the differences materially affects the margin, or if a reasonably accurate adjustment can remove the effect.

That is why a search has two parts: a quantitative screen that applies measurable criteria to every company in the database in the same way, and a qualitative review of each surviving company’s functions, products and circumstances.

Adjustments

Where differences can be measured, they can be adjusted for rather than used to reject a company:

  • Working capital. Differences in receivables, payables and inventory as a share of sales or cost, multiplied by a suitable interest rate, adjust each comparable’s margin to the tested party’s working capital position.
  • Capacity utilisation. Where a comparable’s fixed costs are under-recovered for reasons the tested party does not share, an adjustment may be justified if it can be measured reliably.
  • Risk. A captive bearing minimal market and credit risk is sometimes compared with entrepreneurs; adjustments here are harder to support and should be explained carefully.

Adjustments are considered in an Arm’s Length Analysis where the facts call for them, and every adjustment is shown separately so it can be reviewed.

Multiple-year data

Comparables are identified on data for the current year. Where the most appropriate method is the resale price method, cost plus or TNMM, the results can be combined with the two preceding years on a weighted average basis. If current-year data is not available when the return is filed, the preceding years can be used, and current-year data that becomes available by the time of assessment is then taken into account.

TP Bharat builds a three-year window ending with the year you choose. Each company’s margin is the sum of its numerators divided by the sum of its denominators over the years with data. You decide how many years a company must have, and whether the final year is required.

The range

Where the method is CUP, resale price, cost plus or TNMM and there are six or more comparables, the range runs from the 35th to the 65th percentile. Percentiles are found by position in the sorted list: multiply the number of companies by the percentile; if the answer is a whole number, take the average of the value at that position and the next one, otherwise use the value at the next higher position. A price inside the range is accepted. A price outside it is replaced by the median.

For regional and global documentation we also give the OECD interquartile range (25th to 75th percentile, linear interpolation) on the same accepted set.

Fewer than six comparables

With fewer than six comparables, or where the profit split method or another method applies, the arm’s length price is the arithmetic mean. The transaction price stands if the arm’s length price differs from it by no more than the notified tolerance: 1% for wholesale trading and 3% in other cases in recent notifications. Wholesale trading here means trading where purchases of finished goods are at least 80% of the total cost of the activity and average monthly closing inventory is no more than 10% of its sales.

The tolerance is tested on price, not margin. For a cost-based PLI the price is cost × (1 + margin); for a sales-based PLI it is cost ÷ (1 − margin); for the Berry ratio the price moves in proportion to the ratio.

  1. Characterise the tested party. Its functions, assets and risks decide the activity, the PLI and the filters.
  2. Choose the data and the years. TP Bharat Detailed for the cleanest margins, TP Bharat Abridged for wider coverage.
  3. Choose the industry. Industry groups, main products and services and NIC codes, with keywords on business descriptions.
  4. Apply quantitative filters to all companies in the same way, recording how many each one removes.
  5. Review each surviving company and record the decision and the reason.
  6. Compute the range on the accepted set and compare the tested party.

Steps 2 to 4 are what the free search does. Steps 5 and 6, and the documentation of all six, make up the analysis.

Quantitative filters

FilterWhy it is usedCommon practice
Minimum turnover (required)Very small companies do not operate under comparable conditions, and their accounts are less reliable.Operating revenue of at least ₹1 crore in every year used: the TPO standard, upheld by tribunals (Genisys Integrating Systems, Bangalore ITAT). Applied to every TP Bharat search.
Positive net worth (required)A company whose net worth is eroded is in distress, not in normal business conditions (Rule 10B(3)).Net worth must be reported and positive in the latest year reported. Applied to every TP Bharat search.
Turnover bandScale changes cost structures, bargaining power and risk.A band around the tested party, often one-tenth to ten times its turnover. Upper limits for very large companies are common.
Persistent lossesLosses in most years point to start-up, distress or unusual conditions rather than a normal return.Exclude companies with losses in two of three years, or in the final year.
Related-party transactionsA comparable should be independent. Large related-party dealings can distort its margin.Exclude above 15% to 25% of revenue or cost, in any year of the window.
Activity shareThe comparable should earn most of its revenue from the same kind of activity.75% or more from services, manufacturing or trading, as the case may be.
Export earningsExport-oriented service providers face different markets and risks from domestic ones.25% or more of revenue for IT, ITeS and KPO service providers (75% for captives).
Employee costLabour-intensive services should be compared with labour-intensive businesses.25% or more of operating cost for service providers.
R&D, selling, royaltyHigh spend suggests intangibles that a routine tested party does not own.Commonly 3% of revenue for R&D and selling, with a similar limit for royalties paid.
Abnormal items, falling revenue, year endExceptional or extraordinary items, revenue falling in every year, and a financial year that does not end on 31 March (Rule 10B(4)) make a year unrepresentative.Abnormal items at most 5% of revenue; companies with diminishing revenue and non-March year ends excluded.
Super-normal profitA margin far outside the rest of the set points to something other than the routine activity. Super-normal profit is a reason to examine the comparable (Agnity India, Delhi HC), not an automatic exclusion (Chryscapital, Delhi HC).Exclude far outliers: beyond Q3 + 3 × IQR or below Q1 − 3 × IQR of the set, when at least 20 companies remain; or outside chosen percentiles.
Period, ownership, data checksAccounts for other than twelve months, government control, figures that do not add up or do not match between statements, and sources disagreeing on revenue can make results unrepresentative.Excluded unless there is a reason to keep them.
Filters need reasons. Every one of these filters has been accepted in some cases and rejected in others by the Tribunal. Apply them to all companies in the same way, keep a record of what each removed, and be ready to explain the threshold. The funnel in the search and the screening sheet in the workbook keep that record.

Qualitative review

The review asks whether each company actually does what the tested party does. We read the business description, the products and services, the annual report and any segment information. Companies with significant intangibles, a merger, demerger or restructuring in the window, a different accounting period, or clearly different functions are rejected, and the reason is recorded against each one.

Documentation and forms

  • Local documentation is required where international transactions exceed ₹1 crore in the year (₹20 crore for specified domestic transactions), including the comparability analysis and the search.
  • Accountant’s report in Form 48 (previously Form 3CEB), due by 31 October following the tax year, one month before the return due date of 30 November.
  • Master file in Form 56 for groups with consolidated revenue above ₹500 crore and international transactions above ₹50 crore (₹10 crore for intangibles), due by 30 November; intimation in Form 57.
  • Country-by-country report in Form 59 for groups with consolidated revenue above ₹6,400 crore, within twelve months of the end of the parent’s reporting year.

Failure to keep or furnish documentation can cost 2% of the value of each transaction. Under the 2025 Act a late accountant’s report attracts a fee of ₹50,000 if it is up to a month late and ₹1 lakh after that.

Certainty options

  • Safe harbour. From tax year 2026-27 the IT services categories (software development, IT-enabled services, KPO and contract R&D in software) are combined, with a single 15.5% mark-up on operating cost for transactions up to ₹2,000 crore, and the option can cover five years. Other categories, such as auto components, intra-group loans, guarantees and the new data centre category, have their own conditions.
  • Advance pricing agreement. Fixes the method or price for up to five future years, with rollback to four earlier years. Under the 2026 Rules, rollback is available only if the return for the year was filed by the original due date.
  • Multi-year arm’s length price. From tax year 2026-27, an arm’s length price determined for a transaction can be applied to the next two years if the transaction is materially the same (option in Form 46, certificate in Form 47).
  • Mutual agreement procedure. Resolves double taxation under a treaty. It is not available once a safe harbour has been accepted.

The limits of a search

A quantitative search is only as good as the data and the filters behind it. Recent years fill up gradually as annual reports are filed, industry classifications can be broad, and some functional differences only show up in an annual report. The free search is meant to show whether a credible set exists and roughly where its range sits. The review is what makes it reliable.

Try a search on your own case

It is free and you do not need an account. Company names stay hidden until you order the analysis.