How to choose a transfer pricing method you can defend.

01.What the best method rule actually requires

The U.S. regulations state the standard plainly at §1.482-1(c). There is no priority ordering of methods. The method that governs is the one that, under the facts and circumstances, provides the most reliable measure of an arm's length result, and reliability is judged principally on two factors: the degree of comparability between the controlled and uncontrolled transactions, and the quality of the data and assumptions used.

Two features of that formulation are routinely overlooked. The first is that it is comparative. You cannot establish that a method is the most reliable without saying what it was more reliable than. A section that reaches a conclusion without a comparison has not applied the rule, whatever it says in its final sentence.

The second is that reliability is a property of the data, not of the method. There is no such thing as a reliable method in the abstract. A CUP analysis built on one loosely similar transaction is less reliable than a TNMM analysis built on a well-screened set, and the regulations are indifferent to the fact that the CUP sits higher in most practitioners' mental hierarchy.

The OECD standard at paragraph 2.2 of the Guidelines is substantively parallel: select the method most appropriate to the circumstances of the case. The OECD does express a preference the U.S. regulations do not, stating that where a traditional transaction method and a transactional profit method can be applied with equal reliability, the traditional method is preferred, and that where a CUP and another method can both be applied reliably, the CUP is preferred. That preference is conditional on equal reliability, which is the condition most studies never test.

What follows is the sequence I use. Five questions, in order, each one narrowing the field. The order matters: answering them out of sequence is how a study arrives at a method before it has established what it is pricing.

Step one: What is the transaction you are actually pricing?

Method selection cannot begin until the transaction is delineated. Chapter I, Section D.1 of the Guidelines sets out the exercise: identify the commercial or financial relations between the parties, and the conditions and economically relevant circumstances attaching to them, by reference to the contractual terms, the functions performed, the assets used and the risks assumed, the characteristics of the property or services, the economic circumstances, and the business strategies.

The output is a description of what is being transferred and on what terms. Until that exists, any method is being applied to an assumption.

Delineation changes method selection in a way practitioners underestimate. A supply arrangement that the contract calls a sale of finished goods may, on the evidence of conduct, be a sale of goods plus a licence of marketing intangibles plus the provision of technical services. Three transactions, potentially three methods. Bundling them and running a single TNMM is a decision, and one that should be stated and defended rather than made silently by the shape of the report.

The test: write one sentence describing the transaction without naming a method. If that sentence contains two verbs that belong to different arrangements, you have more than one transaction.

Step two: Can a traditional transaction method be applied?

The three traditional methods compare prices or gross margins directly, and where the data supports them they are the most direct measure available.

The CUP method compares the price charged in the controlled transaction with the price in a comparable uncontrolled transaction. An internal CUP, where the group sells the same product to independent customers on comparable terms, is the strongest evidence in transfer pricing and the most commonly overlooked. External CUPs are harder, but in commodity trades a published quotation can serve as the reference price, with adjustments for quality, delivery terms, volume and the date of price determination. The OECD's commodity guidance in Chapter II addresses this directly, and examiners in commodity-exporting jurisdictions look for it specifically.

The resale price method works back from the price at which a product bought from a related party is resold to an independent, deducting an arm's length gross margin. It fits a distributor that adds no significant intangible value.

Cost plus adds an arm's length gross mark-up to the supplier's costs. It fits contract manufacturing and routine service provision.

Both gross-margin methods share a vulnerability worth confronting early. Gross margin depends on where a company draws the line between cost of sales and operating expenses, and that line moves between companies and between accounting frameworks. Unless the comparables' disclosure supports a consistent classification, a resale price or cost plus analysis is measuring an accounting convention rather than an economic return.

"There is no such thing as a reliable method in the abstract. Reliability is a property of the data, not of the method."

Step three: Is one side of the transaction clearly the simpler one?

This question decides whether a one-sided profit method is available at all.

TNMM, and its U.S. analogue the comparable profits method at §1.482-5, examines the net profit indicator a taxpayer realises from a controlled transaction relative to an appropriate base — costs, sales, or assets. It is one-sided: you test the less complex party and leave the residual with the other side.

That simplification is sound only where one side genuinely is less complex. The tested party should perform routine functions, own no non-routine intangibles, bear no economically significant risk it does not control, and have reliable comparables available. Where that description fits, TNMM is often the most reliable method available, and its tolerance of accounting differences is a genuine advantage over the gross-margin methods.

Where it does not fit, applying TNMM anyway assigns the entire residual to whichever entity you chose not to test. That is not a conclusion the analysis supports; it is a consequence of how the analysis was set up.

Step four: Do both sides make unique and valuable contributions?

If the answer is yes, the one-sided methods are unavailable and a profit split is indicated.

Chapter II, Part III, Section C identifies the circumstances where a transactional profit split is likely to be the most appropriate method: where each party makes unique and valuable contributions, where the operations are highly integrated, and where the parties share the assumption of economically significant risks or separately assume closely related risks. The U.S. equivalent sits at §1.482-6, including the residual profit split, which first assigns routine returns to each party and then divides the residual by reference to the non-routine contributions.

Profit splits are under-used relative to the fact patterns that call for them, and the reason is practical rather than technical. They require data from both sides, a defensible allocation key, and a documentation standard materially harder to meet. None of that makes a one-sided method correct where the facts point elsewhere. It makes the profit split expensive, which is a different problem.

Step five: Which surviving method has the better data?

By this point the field is usually down to two. The last question is empirical, and it is where the comparison actually happens.

For each surviving method, set out what data it would require, what data exists, and what adjustments the gap demands. A CUP with one marginally comparable transaction and three unquantifiable differences is less reliable than a TNMM with eleven well-screened comparables, and saying so explicitly — with the comparable counts, the screening criteria and the adjustment reasoning on the page — is a stronger position than asserting a preference.

This is also where to record the methods you rejected and why. The U.S. documentation rules at §1.6662-6(d)(2)(iii)(B) require a description of the alternative methods considered and an explanation of why they were not selected, and that requirement is a condition of the penalty protection the documentation exists to provide. A method selection section that names no alternatives may cost a taxpayer that protection even where the result itself is unimpeachable.

The test: for the method you rejected, could a reader reconstruct your reasoning from the report alone? If the only stated reason is that no comparables were found, say where you looked.

02.Six signs a method selection won't survive review

These are the patterns that draw questions. Each one is cheap to fix before filing and expensive to explain afterwards.

• No alternative is named. The chosen method is announced rather than selected, with no record of what else was considered.
• The reasoning contains no facts about the business. The section would read identically for any taxpayer in the industry, which means it was drafted from a template rather than from the functional analysis.
• Internal transactions were never evaluated. The group sells to or buys from independent parties on comparable terms, and no one assessed whether that produces an internal CUP.
• The tested party owns something non-routine. A one-sided method is applied to an entity holding intangibles or bearing entrepreneurial risk, which quietly allocates the residual elsewhere.
• A gross-margin method rests on unexamined accounting. Resale price or cost plus is applied without confirming that the comparables classify cost of sales consistently.
• The residual lands nowhere anyone can explain. Apply the method across the group and a substantial profit settles in an entity with few people and little activity.

03.A final word

None of this is an argument against TNMM, which remains the appropriate method for most controlled transactions for good reasons. It is an argument against arriving at TNMM without having considered anything else, because that route produces two distinct failures: a documentation gap that is easy for an examiner to identify, and occasionally a substantive misallocation that is expensive to unwind years later.

The remedy costs an afternoon. Before the comparable search begins, write down the methods available, the data each would need, what exists, and the specific reason each is or is not viable on these facts. Most of the time the exercise confirms the method you expected, and you finish with a section that does its job under scrutiny.

Occasionally it will not. Those are the cases the afternoon was for.