That matters because the two commonest documentation failures are opposite ones. Running a TNMM by reflex and never recording why a CUP was unavailable is the first. Treating the TNMM as a method that needs apologising for is the second.
What the Guidelines actually say?
Para. 2.1 of the OECD Transfer Pricing Guidelines names five methods in two families. CUP, resale price and cost plus are the traditional transaction methods; the TNMM and the transactional profit split are the transactional profit methods.
Para. 2.2 sets the standard — the selection "always aims at finding the most appropriate method for a particular case" — and the criteria: the strengths and weaknesses of the recognised methods, the appropriateness of the method given the nature of the transaction as established by the functional analysis, the availability of reliable information on uncontrolled comparables, and the degree of comparability including the reliability of any adjustments needed. It closes with the sentence that saves the most work: "No one method is suitable in every possible situation, nor is it necessary to prove that a particular method is not suitable under the circumstances."
Para. 2.3 then adds two tie-breaks, and only for the case where two methods really are equally reliable: the traditional method is preferable, and the CUP "is to be preferred". Para. 2.12 removes the belt-and-braces habit — the arm's length principle "does not require the application of more than one method for a given transaction".
The rule is therefore narrow. You do not eliminate four methods. You select one and say why it fits the transaction and the data you actually have.
Which method fits which transaction?
Para. 2.64 defines the fourth of these: the TNMM "examines the net profit relative to an appropriate base (e.g. costs, sales, assets) that a taxpayer realises from a controlled transaction". Note that it does not say "distributor margin". The base follows the facts.
Why the TNMM wins anyway?
Criterion three settles it. Comparable uncontrolled prices for a specific intercompany transaction rarely exist outside commodity and financial markets; net profit indicators for independent companies with comparable functions do, in published accounts, in quantity. Para. 2.75 explains why that is more than convenience: "Prices are likely to be affected by differences in products, and gross margins are likely to be affected by differences in functions, but net profit indicators are less adversely affected by such differences."
The practice data agrees, with a caveat. The US IRS advance pricing agreement reports are the only series publishing a method breakdown: across the three most recent years they put the CPM/TNMM at 78–86 % of tangible and intangible property transactions and 83–91 % of services transactions, most recently Announcement 2026-8 of 30 March 2026. That is US data — a directional indicator for European work, not a local benchmark. We are not aware of any EU-level statistic breaking method use down by percentage; the Joint Transfer Pricing Forum's APA statistics do not report method at all, and its October 2016 report on comparables notes only that external data serve resale minus, cost plus "and in particular the Transactional Net Margin Method".
Nor did the EU choose to codify a preference. The proposed Transfer Pricing Directive, COM(2023) 529, required the most appropriate method and said in Recital 10 that it had no preference among the recognised methods; the Commission withdrew it on 21 October 2025. The one place a method is now effectively prescribed is Amount B, which supplies a return-on-sales result from a fixed matrix for in-scope baseline marketing and distribution in adopting jurisdictions, for fiscal years beginning on or after 1 January 2025 — a TNMM with the answer filled in. Where it does not apply, the transfer pricing benchmarking study still sets the number. Our note on Pillar Two and transfer pricing documentation covers the other direction from which minimum-tax rules now read the same file.
Where the TNMM is the wrong answer
Intercompany financing. Interest rates have an observable market, so arm's length interest rates are a CUP exercise under Chapter X rather than a net-margin proxy.
Both parties contribute something unique. If neither side has the less complex functional analysis, the TNMM has no tested party to sit on, and a profit split is the honest answer.
A clean cost base and a genuinely routine function. Cost plus can beat a TNMM on reliability — but both halves have to be built, the base and the rate. In its judgment of 9 August 2023 (I R 54/19) the German Federal Fiscal Court left cost plus for a toll manufacturer undisturbed and faulted everything downstream of it: principal-supplied materials bought back after processing were held not to be value-adding costs belonging in the base, and a mark-up drawn from general experience and internet research was held insufficiently substantiated. The case was remitted for the rate to be determined.
Then the quiet failure. Para. 2.12 excuses you from testing a second method, not from the selection: a TNMM chosen because the data was easy, with nothing on file about why a CUP or cost plus was rejected, is a documentation gap rather than a method error. The transfer pricing methodology section of the file is where that reason belongs.
What decides a TNMM result?
The tested party. Para. 3.18: it "will most often be the one that has the less complex functional analysis".
The profit level indicator. Cost-based indicators are appropriate only where costs are a relevant indicator of the value of the functions performed, assets used and risks assumed (para. 2.98); Berry ratios, defined at para. 2.106, only under the conditions at para. 2.107.
Consistency of measurement. The indicator must be computed identically for the tested party and every comparable — accounting differences, not economics, decide most arguments.
The range. Where comparability defects remain that cannot be identified or quantified, para. 3.57 allows statistical tools such as the interquartile range — they "might assist in enhancing the reliability of the analysis", which is permissive rather than the automatic trim many studies apply to the output of a comparable companies analysis. Para. 3.60 is the relief: inside the arm's length range, no adjustment. Para. 3.62 is the exposure: where one is required, the median may be used. As our note on proportionality sets out, the effort spent finding that range should match the transaction rather than be maximal.
What this means for the data?
If the method turns on the availability of reliable information, the data set is what decides which method you are entitled to use. Two properties carry that, and they pull the same way.
Coverage of the right population. Genuine comparables for a routine European entity are overwhelmingly private companies, and listed-only sets are too thin to support a TNMM twice. Benchmarking Pro, the transfer pricing database smartZebra launched on 27 July 2026, is built for that: more than 500,000 curated company data points across 15+ European countries, licensed on a flat fee rather than per search, and included in existing Transfer Pricing Pro plans. Counted across the platform, the comparable population runs to more than 500,000 private companies alongside the 50,000-plus listed profiles — roughly nine in ten of it private.
Traceability. Every ratio has to lead back to a named company's published accounts, with the accounting basis, the period and the rejection reasons on the record. The transfer pricing module covers the ratios TNMM work actually uses — EBIT margin, net cost plus, gross cost plus, adjusted net cost plus, gross margin, return on assets and the Berry ratio — applies quantitative screens for independence, losses and R&D intensity, classifies comparables by SIC and NAICS code, calculates the interquartile range and median, and opens the original source report behind every data point. Transfer Pricing Pro is the transfer pricing software layer that exports the benchmarking appendix into the file. The automation sits in the mechanical steps; the method choice stays with the adviser, which is where the Guidelines put it. What makes such a database defensible in the first place is set out in our guide for tax consultants.
The TNMM is not the default because it is the best method. It is the default because it is usually the only one the data supports — and saying so, on the file, is what makes the choice defensible.
References
- OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations 2022 — Ch. II paras. 2.1, 2.2 (most appropriate method and selection criteria), 2.3 (double preference), 2.12, 2.64 (TNMM definition), 2.75, 2.98, 2.106–2.107 (Berry ratio); Ch. III paras. 3.18 (tested party), 3.57 (interquartile range), 3.60 and 3.62 (position in the range)
- OECD, Pillar One — Amount B: simplified and streamlined approach, February 2024, annexed to Chapter IV of the Guidelines — return-on-sales pricing matrix, effective for fiscal years beginning on or after 1 January 2025 in adopting jurisdictions
- US Internal Revenue Service, Announcement 2026-8, 30 March 2026 (27th Annual APA Report, CY2025: 110 APAs; CPM/TNMM 86 % of tangible and intangible property transactions, 83 % of services), with Announcement 2025-13 (CY2024) and Announcement 2024-16 (CY2023) for the three-year range
- EU Joint Transfer Pricing Forum, Report on the Use of Comparables in the EU, October 2016
- European Commission, Proposal for a Council Directive on transfer pricing, COM(2023) 529 final, 12 September 2023 — Art. 10(1) and Recital 10; withdrawn 21 October 2025 (2026 Commission Work Programme)
- Bundesfinanzhof, judgment of 9 August 2023, I R 54/19 — cost plus for a toll manufacturer: principal-supplied materials excluded from the cost base, mark-up from general experience and internet research insufficiently substantiated; case remitted
Related pages
- Transfer Pricing Benchmarking & TNMM Analysis — the module, the profit level indicators and the calculation log on live data
- Benchmarking Pro — transfer pricing database with 500,000+ company comparables — ratio coverage, country coverage and the flat-fee model
- Transfer Pricing Databases: A Guide for Tax Consultants — selection criteria and what auditors reject
- How Much Search for Comparables is Enough? — the OECD's proportionality standard for benchmarking effort
- Your Benchmarking File Is Now a Pillar Two Input — how the global minimum tax reads the same study
- Credit Spreads & Interest Rates — CUP-based intercompany loan pricing under OECD Chapter X
.png){kind=spacer}
