France Transfer Pricing

Contact

Choosing the Most Appropriate Transfer Pricing Method: A Practical Decision Framework

Written by

Terence WILHELM

in

,

The transfer pricing method is not selected by habit, by industry label or by database availability. It is selected after the controlled transaction has been accurately delineated and the relative reliability of the available methods has been assessed.

This guide provides a structured way to choose among the five OECD-recognised methods, identify the tested party and financial indicator where relevant, and document why the selected method is the most appropriate for the facts.

Core rule. The objective is not to prove that every other method is impossible. The objective is to show that the selected method provides the most reliable measure of an arm’s-length outcome, given the transaction, the functions performed, the information available and the degree of comparability that can reasonably be achieved.

1. Method selection starts with the transaction—not the database

Before a method can be selected, the transaction must be accurately delineated. This requires more than reading the intercompany agreement. The analysis should identify:

  • the contractual terms and the parties’ actual conduct;
  • the functions performed, assets used and economically significant risks assumed and controlled;
  • the characteristics of the goods, services, financial instruments or intangibles;
  • the economic circumstances and markets in which the parties operate;
  • the business strategies and realistic alternatives available to each party.

A method cannot repair a poorly defined transaction. If the transaction perimeter changes during the analysis, the tested party, profit indicator and comparables may all need to change as well.

2. The four selection criteria

Chapter II of the OECD Transfer Pricing Guidelines 2022 identifies four connected considerations:

  1. Strengths and weaknesses of each method. A direct price comparison is powerful when the transactions are sufficiently comparable; a net-margin method is more tolerant of some product differences but can be distorted by functional or accounting differences.
  2. Nature of the accurately delineated transaction. The method must fit the functions, assets and risks of the parties.
  3. Availability of reliable information. This includes information on the controlled transaction itself and on uncontrolled comparables.
  4. Degree of comparability and reliability of adjustments. Material differences must either be absent or capable of reasonably accurate adjustment.

There is no mechanical hierarchy that always produces the answer. However, where a reliable comparable uncontrolled price can be identified, it ordinarily offers the most direct link to the arm’s-length condition. Conversely, the lack of a perfect comparable does not automatically justify a profit split or an untested rule of thumb.

3. The five recognised methods

Comparable uncontrolled price (CUP)

The CUP method compares the price or rate charged in a controlled transaction with the price or rate in a comparable transaction between independent enterprises. A comparable may be internal—one of the parties transacts with an independent customer or supplier—or external.

  • Often suitable for: commodities, standardised products, observable financing transactions, licences or services where genuinely comparable terms exist.
  • Key sensitivities: product characteristics, volumes, markets, contractual allocation of risk, delivery terms, credit quality, currency, duration and embedded options.
  • Practical test: can the material differences be identified and adjusted with sufficient reliability?

Resale price method

This method begins with the price at which a product acquired from an associated enterprise is resold to an independent customer. An arm’s-length gross resale margin is deducted to compensate the distributor for its functions, assets and risks, leaving the transfer price.

  • Often suitable for: routine distribution where the reseller adds limited value and does not use unique intangibles.
  • Key sensitivities: consistency of gross-margin accounting, marketing intensity, warranty and inventory risks, level of market, and differences in operating structures.
  • Common weakness: apparently comparable distributors may classify costs differently between cost of sales and operating expenses.

Cost plus method

The cost plus method adds an arm’s-length gross mark-up to the direct and indirect costs incurred by the supplier. The cost base and the mark-up must both be defined consistently.

  • Often suitable for: routine manufacturing, semi-finished goods and service activities where the provider performs limited functions and assumes limited risks.
  • Key sensitivities: the composition of the cost base, pass-through costs, capacity utilisation, shareholder costs, exceptional expenditure and accounting consistency.
  • Common weakness: applying a market mark-up to a cost base that does not correspond to the base used by the comparables.

Transactional net margin method (TNMM)

The TNMM compares the net profit earned by a tested party from the controlled transaction with the net profit that independent enterprises earn from comparable activities. The result is measured against an appropriate base such as sales, costs or operating assets.

  • Often suitable for: a less complex distributor, manufacturer or service provider for which reliable net-margin comparables exist.
  • Key sensitivities: selection of the tested party, transaction-level segmentation, working-capital differences, tested operating expenses, loss-making comparables and multi-year data.
  • Common weakness: using entity-wide results when the company performs several materially different activities.

Transactional profit split method

The profit split method identifies the relevant combined profit or loss from the controlled transactions and allocates it between the associated enterprises using economically valid factors that approximate the division independent parties would have agreed.

  • Potential indicators: both parties make unique and valuable contributions; their operations are highly integrated; or they share economically significant risks, or assume closely related risks.
  • Key sensitivities: definition of the relevant profit, consistent accounting, treatment of routine contributions, choice and measurement of splitting factors, and whether actual or anticipated profits should be split.
  • Important caution: the absence of close comparables, by itself, does not make profit split the most appropriate method.

4. Choosing the tested party for a one-sided method

The choice of a tested party is not confined to the TNMM. It arises whenever a one-sided method tests the price or margin of one participant rather than the combined outcome of both parties. The resale price method, the cost plus method and the TNMM all require the analysis to be anchored to the participant—and to the relevant transaction or reliably segmented activity—to which the method can be applied most reliably.

The preferred side is generally, but not mechanically, the one with the less complex functional profile. The fewer the unique contributions, controlled risks and distinctive assets on that side, the easier it is to identify independent enterprises performing sufficiently comparable activities and to make reliable adjustments. The practical question is therefore not which entity has the most convenient accounts, but which participant’s contribution can be isolated and benchmarked with the least uncertainty.

The analysis must remain transaction-specific. An entity may be the simpler party for one controlled transaction and not for another. A company that owns unique intangibles, controls significant risks or performs highly integrated functions is unlikely to be an appropriate tested party for a routine return. Where both parties make unique and valuable contributions or their activities cannot be reliably separated, a one-sided method may be inappropriate and a profit split may need to be considered.

5. Selecting a replicable margin indicator

Margin-based methods may examine a gross margin or a net margin. The relevant indicator must reflect the economics of the tested activity: the resale price method focuses on a gross resale margin, the cost plus method on a gross mark-up over an appropriate cost base, and the TNMM on a net profit indicator such as sales, costs or operating assets. In each case, the numerator and denominator must be defined consistently and must not be distorted by the controlled transaction being tested.

A taxpayer can design an indicator that closely reflects its own value drivers, but the decisive practical constraint is replicability. The same calculation must be capable of being reconstructed for the independent comparables. If the taxpayer’s accounts provide a refined aggregation that commercial databases do not disclose—or if the comparables classify costs differently—the same indicator cannot be reproduced on both sides. The result is a break in comparability, even if the indicator is economically persuasive in the abstract.

  • Gross resale margin: may be suitable for routine distribution, provided cost-of-sales classification is sufficiently consistent between the tested activity and the comparables.
  • Gross mark-up on costs: may be suitable for manufacturing or service activities, but the cost base, pass-through items and accounting treatment must be capable of consistent reconstruction.
  • Operating margin on sales: is often used for routine distributors where sales are the relevant value driver and operating expenses can be compared reliably.
  • Net cost plus: may be appropriate for routine services or manufacturing if the tested cost base and the comparable companies’ cost bases are aligned.
  • Berry ratio or other value-added indicators: may be considered for limited intermediary functions, but only if the denominator captures the relevant contribution and the database provides the necessary aggregates.
  • Return on operating assets: may be relevant for asset-intensive activities if asset measurement, ownership and utilisation are comparable.

Before adopting an indicator, the analyst should therefore test three bridges: it must follow from the functional analysis, it must be calculable from reliable segmented data for the tested activity, and it must be reproducible from the financial information available for the comparables. Failure at any one of those stages weakens the method.

Method selection and Article 57 exposure

Method selection also has a practical evidentiary dimension in France. The transfer pricing documentation describes the taxpayer’s own analysis and may frame the subsequent audit discussion. Article 57 is drafted broadly: an indirect transfer of profits may arise through increased or reduced purchase or sale prices or by any other means. In addition, for financial years opened on or after 1 January 2024, where the method actually applied departs from the method described in the documentation, the resulting difference is presumed to be an indirect transfer of profits unless the taxpayer demonstrates the absence of a transfer. A method that displays persistent losses or an unusually low return for the French tested party may therefore create immediate exposure, particularly if the documented functional profile presents that party as routine or low-risk.

This does not justify choosing a method solely because it produces a more favourable result. It does require the taxpayer to investigate whether the weak outcome reflects commercial circumstances, an unreliable segmentation, an unsuitable indicator or a mismatch between the selected method and the controlled transaction. Where a gross-margin method more directly isolates the controlled pricing and can be implemented with sufficiently consistent accounting data, it may provide a more relevant first line of analysis than a TNMM that mechanically places a weak or negative net result at the centre of the file. Its greater implementation difficulty must nevertheless be acknowledged and documented.

A robust method-selection file should therefore stress-test both gross and net perspectives where they are realistically available, explain any divergence and show why the final method remains the most appropriate rather than merely the most convenient.

6. Internal comparables: powerful evidence, but not automatically conclusive

Internal comparables should be examined before an external database search. They are not limited to an internal CUP. Transactions entered into by the taxpayer—or, where sufficiently documented, by another group entity—with independent customers or suppliers may provide evidence of prices, gross margins or net margins. Internal data can therefore support the CUP, resale price, cost plus or TNMM, depending on the transaction and the financial information that can be reliably segmented.

  • Price evidence: the same or a similar product, service, licence or financial instrument is supplied to or obtained from an independent party.
  • Gross-margin evidence: comparable third-party distribution, manufacturing or service transactions allow a resale margin or cost mark-up to be reconstructed.
  • Net-margin evidence: reliable segmentation makes it possible to compare the net return earned on comparable controlled and uncontrolled activities.

Differences between the third-party and intra-group transactions are common and do not make the internal reference irrelevant. In dealings with independent parties, the taxpayer may perform additional commercial or negotiation functions, assume greater market, credit, inventory or contractual risks, use different assets, or provide a broader package of services. By contrast, an intra-group participant may operate within a more integrated framework and perform fewer functions or assume fewer risks. Those differences are economically plausible, but they must be demonstrated rather than presumed.

The analysis should identify each material difference and determine whether a reasonably accurate adjustment is possible. Relevant factors may include volumes, geography, contractual duration, credit terms, product or service scope, allocation of risk, marketing intensity, exclusivity, strategic market entry and the level of integration between the parties. Where the differences are too numerous or cannot be measured reliably, the internal reference should not be presented as a stand-alone comparable.

In practice, an internal comparable will rarely be self-sufficient unless the uncontrolled and controlled transactions are closely aligned and any material differences can be reliably adjusted. It may nevertheless provide valuable corroboration: it can confirm the direction or order of magnitude of an external benchmark, explain why an intra-group margin differs from a third-party margin, or test whether the result produced by the selected method is commercially coherent. Where comparability is genuinely strong, however, an internal comparable may remain the most direct and reliable primary evidence and should not be relegated to a merely secondary role.

7. From a benchmark to an arm’s-length range

A benchmarking study normally produces a distribution of results rather than a single arm’s-length figure. Once the observations are ranked from the lowest to the highest, the first quartile (Q1) is the 25th percentile, the median (Q2) is the 50th percentile and the third quartile (Q3) is the 75th percentile. The interquartile range extends from Q1 to Q3: it retains the central 50% of the observations and excludes the lowest 25% and the highest 25%. The median is the central point of the ordered distribution; unlike the arithmetic mean, it is less sensitive to extreme values.

The calculation is not entirely self-executing. Databases and statistical software may use different interpolation conventions when a quartile falls between two observations. A defensible study should therefore retain the full set of results, identify the convention used to calculate Q1, Q2 and Q3, and apply it consistently. More importantly, statistical refinement cannot cure a weak search strategy or make poorly comparable companies reliable.

The interquartile range is a reliability tool, not a safe harbour

The OECD Guidelines do not prescribe the interquartile range in every case. They contemplate statistical tools such as the interquartile range where a sizeable set of observations still contains unidentified or unquantifiable comparability defects after the less reliable results have been removed. If the relevant condition of the controlled transaction lies within a reliable arm’s-length range, no adjustment should be made. Where the points in that range are of relatively equal and high reliability, any point may satisfy the arm’s-length principle. A measure of central tendency becomes relevant only where residual comparability defects remain and its use reduces the risk of error.

The median is a context-dependent reference, not an automatic legal destination

French administrative guidance states that, in the absence of evidence identifying a more appropriate point, it is customary to regard the median as the remuneration towards which the taxpayer should tend. The same guidance immediately qualifies that statement: any point within the arm’s-length range is acceptable where the study is highly reliable, and the interquartile range is described as a common practice rather than a statutory definition of the arm’s-length range.

That guidance expresses the tax administration’s interpretation. Under Article L. 80 A of the French Tax Procedures Code, published administrative guidance may, subject to the statutory conditions, be invoked by a taxpayer against the administration. It does not create an autonomous transfer pricing rule that binds taxpayers and courts beyond the legislation it interprets. The median must therefore not be confused with a regulatory safe harbour.

French case law illustrates this fact-sensitive approach. In TCL Belgium, the Versailles Administrative Court of Appeal held that the median could not, by itself and as a matter of principle, constitute the relevant reference for the admissible range. It cancelled the adjustment for a year in which the taxpayer’s margin was within the interquartile range and, for another year, limited the adjustment by reference to the lower bound of that range. In General Electric Medical Systems, the Conseil d’État subsequently accepted the median because the court below had found that, in the circumstances of the case, it was the point that best reflected the facts and circumstances of the transactions. The decision did not establish a universal median rule. More recent decisions remain fact-dependent: for example, the Lyon Administrative Court of Appeal upheld a median-based adjustment where the administration had not applied the median automatically, the taxpayer’s margins were below even the extreme minima of the sample and the taxpayer had not established circumstances supporting another point.

Practical conclusion. The construction of the range, the finding that a controlled result falls outside it and the selection of the point used to quantify an adjustment are three separate questions. The better reading is not that the median is a default legal target which the taxpayer must always disprove. The point selected—particularly by the tax authority when quantifying an adjustment—should be connected to the quality of the comparables and to the facts and circumstances of the transaction. A taxpayer relying on another point should nevertheless document why that position reflects its functions, risks, market conditions and degree of comparability more reliably.

  • preserve the full distribution and the quartile calculation convention;
  • explain why the interquartile range improves reliability rather than applying it mechanically;
  • distinguish the test for identifying a transfer of profits from the choice of the point used to quantify any adjustment;
  • where the tested result lies within a reliable range, document why no adjustment is warranted;
  • where the tested result lies outside the range, explain why the selected adjustment point—not merely the median as a label—best reflects the transaction.

8. A practical decision sequence

  1. Delineate the transaction. Identify the actual conduct, value chain, risks, assets and realistic alternatives.
  2. Test for reliable internal comparables. If they exist, assess whether a CUP or another traditional transaction method can be applied.
  3. Consider traditional methods. Determine whether price or gross-margin comparability is sufficiently reliable.
  4. Identify whether one party is less complex. If so, assess a one-sided method and justify the tested party.
  5. Select the financial indicator. Link it to the tested activity and establish reliable segmented results.
  6. Assess profit split indicators. Use profit split because the contributions and integration justify it—not merely because comparables are scarce.
  7. Stress-test the result. Review consistency with contracts, budgets, actual conduct, commercial outcomes and other transfer pricing positions.
  8. Document the alternatives considered. A concise explanation of why the chosen method is more reliable is generally more useful than a formulaic rejection of every method.

9. Transaction-specific orientation

  • Tangible goods: begin with possible CUPs; otherwise examine resale price, cost plus or TNMM according to the functions of the distributor or manufacturer.
  • Intra-group services: define the benefit, cost base and allocation key before testing the mark-up; distinguish shareholder activities and pass-through costs.
  • Financing: delineate the instrument, borrower creditworthiness, group support, covenants, currency, term and options before selecting a comparable rate or another recognised approach.
  • Intangibles: understand the legal rights and the functions relating to development, enhancement, maintenance, protection and exploitation; do not assume that legal ownership alone determines the return.
  • Business restructurings: analyse what is transferred, options realistically available and whether compensation would be agreed between independent parties.

10. What the French documentation must show

For companies within Article L. 13 AA of the French Tax Procedures Code, the local file must identify the most appropriate method for each material category of controlled transactions and explain why it was selected. Where relevant, it must identify the tested party, state the important assumptions, describe the comparables and search strategy, explain adjustments and reconcile the financial data to the annual accounts.

French administrative guidance similarly emphasises that the method must be justified, consistent with the functions performed and risks assumed, and produce an arm’s-length remuneration. See BOI-BIC-BASE-80-10-10.

Common errors

  • selecting TNMM before defining the transaction because database data are available;
  • calling an entity “routine” without examining who actually controls the risks;
  • combining transactions that have different economics, counterparties or methods;
  • testing an entity-wide margin without reliable segmentation;
  • using a gross-margin method despite inconsistent cost classification;
  • treating the interquartile range or its median as an automatic safe harbour;
  • using a second method mechanically rather than as genuinely useful corroboration;
  • allowing the documentation method to diverge from the method actually implemented.

Expert perspective

The best method is usually the method that survives three reconciliations: with the operational reality, with the accounting data and with the available independent evidence. When one of those bridges is missing, greater methodological sophistication rarely cures the weakness. A clear, proportionate and repeatable method is generally more defensible than a complex model that the business cannot actually operate.

Primary sources

Current at 15 October 2026. This guide provides general information only and does not constitute legal or tax advice. The most appropriate method depends on the accurately delineated transaction and the evidence available in the specific case.

More posts