If your group prices flows between a Romanian entity and its affiliates, each flow must prove it respects the arm's length standard — but no two flows use the same method. The right method is not chosen by preference; it follows from the nature of the transaction and the data available. Three group companies, three different situations, three different methods.
All the methods serve a single idea, enshrined in article 11 of the Romanian Fiscal Code: a transaction between related parties must produce the same result as if it had been concluded between independent companies — the arm's length principle. The methods are only different techniques for measuring this "as if": some compare the price directly, others compare the profit margin, others split the profit between the parties. The Fiscal Code refers to the OECD Guidelines, which describe five methods. Choosing the one "most appropriate" to the nature of the transaction is itself a technical decision — and the first line of defence in an audit. The method is documented in the transfer-pricing file.
Compares the price in the transaction with the affiliate to the price in a comparable transaction between independents. It works excellently where there is a clear reference price (exchange-traded commodities, standardised products, the interest on intra-group loans). The weakness: the smallest difference in volume, delivery, market or timing can make two prices non-comparable without adjustments.
Starts from the price at which the distributor resells the goods to an independent customer and deducts an arm's length gross margin. What remains is the "market" price at which the goods should have been bought from the affiliate. Typical for distributors that do not transform the product but only resell it.
Starts from the supplier's costs and adds an arm's length profit mark-up. Suited to contract manufacturing (where the producer does not bear major market risks) and to intra-group services with routine functions. The key: correctly defining the cost base and the mark-up.
Compares a net profitability indicator (for example, the operating margin) of the tested party with that of comparable independent companies. The most used method: it requires not product comparables but profitability comparables, available in databases. Which is precisely why it is also the terrain of most comparability disputes.
Allocates the combined profit of the transaction between the affiliates, according to each one's contribution. Used when both parties bring unique and valuable contributions (each holds important intangibles), and the other methods do not work. The most complex and assumption-sensitive, but sometimes the only one that reflects economic reality.
The five methods are not a list from which you tick a preferred option, but a toolbox from which you choose the right tool. The OECD also recognises the use of other methods, where none of the five fits, provided the result respects the arm's length principle and is documented. In practice, the same company frequently uses different methods for different flows: TNMM for routine distribution, cost plus for support services, and CUP for the interest on an intra-group loan. Coherence of the file does not mean one method for everything, but the right method for each transaction, with the choice justified.
The taxpayer does not choose the method by whichever result suits it, but by the one most appropriate to the nature of the transaction and the available data. The selection rules take into account:
And the functional profile of the parties — who transforms, who only resells, who holds the intangibles.
Reliable ones, for each method separately.
Attainable, including the adjustments required.
Of each method for the specific situation.
In practice, the OECD order of preference favours the traditional methods (CUP, resale price, cost plus) where the data allow, but recognises that, for many transactions, TNMM is the most practicable. What is not accepted is choosing the "convenient" method — the one that gives the desired figure — instead of the correct one.
No method is chosen before the functional analysis — the description of what each party actually does: the functions performed (production, distribution, marketing, research), the assets used (equipment, trademarks, patents, capital) and the risks assumed (market, inventory, credit, currency). The functional analysis establishes each entity's profile — routine or complex — and, implicitly, how much profit it is due. An entity with simple functions and small risks deserves a small, stable remuneration; one that bears the risks and holds the intangibles deserves the residual profit.
Here appears the close link with economic substance: the functional profile invoked in the file must correspond to reality — to the actual people, decisions and risks. A company that claims routine functions in the file but in fact takes the group's strategic decisions has a false functional analysis, easily dismantled at audit. A method built on a non-compliant functional analysis falls with it — the profile must be supported by real substance.
The methods that look at the margin of a single party (resale price, cost plus, TNMM) require establishing the tested party — the entity whose remuneration is analysed. The rule: you test the party with the simpler functional profile, easier to compare, usually the one that does not hold unique intangibles. In a manufacturer–distributor relationship, the routine distributor is frequently tested; in a service provision, the provider. If the tested party is a foreign group entity, the new framework requires additional guarantees as to the reliability of the calculation — up to an independent auditor's report. Choosing the wrong tested party vitiates the rest of the analysis.
There is rarely a single "correct" price. The comparables produce a range of values — and from it, as a rule, the interquartile range is retained: the extreme values are removed and the central zone is kept, bounded by the lower quartile, the median and the upper quartile. If the tested party's indicator falls within this range, the transaction is compliant. If it lies outside it, the risk of adjustment arises — and the tax authority usually brings the result to the median, not to the nearest edge. Hence the amplified financial effect, treated at length in the article on ANAF adjustments. The details of the comparability study and the interquartile range are treated separately.
The method and the tested party are not formalities to complete at the end. They determine the whole result of the file: the same transaction, analysed through cost plus instead of TNMM, or with the distributor tested instead of the manufacturer, can produce completely different "arm's length" margins. That is why the first move of an inspection is often to contest the method or the tested party — if it knocks them down, the rest of the file becomes irrelevant. And if you choose the convenient method instead of the appropriate one, you hand the inspection exactly the lever it needs.
The most frequent error is not one of calculation but of strategy: choosing the method that produces the desired figure, then building a justification around it. The signs are visible to any experienced inspector: a method chosen without explaining why the others were rejected; a thin functional analysis that "fits" the method suspiciously well; comparables selected to support the result. A file built this way holds up until the first serious analysis. The correct method may be less flattering, but it is defensible — and defensibility, not optimism, is what counts at audit and in litigation. ANAF (the Romanian tax authority), at inspection, essentially asks four questions of the method:
Is the method the most appropriate to the nature of the transaction, or did you choose it for the result?
Why were the other methods rejected? The absence of this justification is a risk signal.
Is the tested party correctly chosen, and does its functional profile match reality?
Do the comparables support the range invoked, or is the selection superficial?
When the answers falter, the inspection can propose its own method and its own range, with adjustment to the median. An effective defence does not begin there, but in the file: a correctly chosen method, with the rejection of alternatives justified and a solid functional analysis, turns a technical discussion into ground on which the taxpayer stands firm. How the method is defended in the inspection is treated separately, and in litigation the decisive role belongs to the transfer-pricing expert report.
None in the abstract. The best is the one most appropriate to the specific transaction and the available data. CUP is ideal where there is a direct reference price; resale price, for distributors; cost plus, for contract manufacturers and routine providers; TNMM, where product comparables are lacking but profitability comparables exist; profit split, for integrated operations with intangibles on both sides.
Because it requires not product comparables but net-profitability comparables, available in databases of financial statements, and it is more tolerant of detailed differences between transactions. For many routine distributions and service provisions it is the practicable method. The reverse: being so widespread, it attracts the most disputes over comparables.
The transfer-pricing file must reflect the method most appropriate from the perspective of the arm's length principle, technically justified. It does not matter which method you use internally for management; what matters is that the method documented in the file is correctly chosen and supported by the functional analysis and comparables. A mismatch between operational reality and the method invoked is exactly what an inspection seeks.
The tax authority must give reasons why your method is not appropriate and why its own is. If the initial justification is solid, the inspection's task becomes hard; if it is weak, the ANAF method may prevail, with adjustment to the median. A divergence of method is one of the most frequent and most contestable issues in litigation, where technical expertise decides.
This article is strictly informational and does not constitute legal or tax advice. Individual situations must be analysed on their specific facts. Legislative position reflected: 18 July 2026.
Deadlines run from the moment of communication. A first discussion clarifies what is being alleged, what you must justify and how the defence is built — before an estimation becomes an assessment decision.