Five OECD methods test whether a group’s internal prices are fair: comparable uncontrolled price, resale price, cost plus, the transactional net margin method (TNMM) and profit split. TNMM, known in the US as the comparable profits method, dominates because the margins of comparable companies are available. Profit split is used when both sides contribute something unique.
Why it matters: The method chosen depends on what comparable data actually exists.
Summary: Five OECD methods apply the arm’s length principle: comparable uncontrolled price, resale price, cost plus, the transactional net margin method (TNMM) and profit split. TNMM, the US comparable profits method, dominates because comparable companies’ margins are available; profit split is used when both sides contribute something unique.
- TNMM tests the simpler party with a profit level indicator, such as markup on total costs, against a range of comparables.
- US rules use the interquartile range and adjust results outside it to the median.
- CPM/TNMM was used for 86% of the tangible and intangible transactions in APAs the IRS executed in 2025.
- In the worked example, pricing below the range pays only if the chance of challenge is under 16%.
- Target the median and book a true-up before the year closes.
Applying the arm’s length principle in practice requires an actual methodology for determining what an independent party would have charged. Several standard methods exist, each suited to different types of intercompany transactions.
| Method | How It Works |
|---|---|
| Comparable Uncontrolled Price (CUP) | Directly compares the intercompany price to a comparable price charged between unrelated parties for the same or a very similar product — the most direct method, when a good comparable actually exists |
| Resale Price Method | Starts from the price at which a distributor subsidiary resells goods to an unrelated end customer, then works backward to an appropriate intercompany price by deducting an appropriate gross margin |
| Cost Plus Method | Adds an appropriate profit margin to the supplying entity’s own costs — commonly used for intercompany manufacturing or service arrangements |
| Transactional Net Margin Method (TNMM) | Compares the overall net profit margin earned on the controlled transaction to margins earned by comparable independent companies — the most commonly used method in practice, given the difficulty of finding a perfect CUP comparable for many transactions |
| Profit Split Method | Divides the combined profit of the related parties according to each side’s contribution, typically by first giving each a routine return and then splitting the residual in proportion to the value of its unique contributions; used when both sides contribute unique intangibles or the operations are too integrated to test one side alone |
The first three are traditional transaction methods, which test prices or gross margins; TNMM and profit split are transactional profit methods, which test operating profit. The US regulations use slightly different names: the comparable profits method (CPM) is the US counterpart of TNMM, and the comparable uncontrolled transaction (CUT) method is the CUP for intangibles. Neither system imposes a strict order. The OECD asks for the “most appropriate method” and the US for the “best method,” meaning the one that gives the most reliable measure on the facts. In practice the one-sided margin method dominates: in the IRS’s report on APAs executed in 2025, CPM/TNMM was used for 86% of the tangible-property and intangible-property transactions covered.
TNMM tests one party, the tested party, which should be the simpler entity (a contract manufacturer or distributor) because reliable comparables exist for simple functions. Its result is measured by a profit level indicator: operating margin on sales for a distributor, markup on total costs for a manufacturer or service provider. Comparable independent companies’ results form a range. In the US the range is usually the interquartile range, the 25th to 75th percentile of the comparables; a result inside it is accepted, and a result outside it is adjusted, ordinarily to the median (Treasury Regulation 1.482-1(e)).
The profit split method is the fifth OECD method and the right tool when both sides make unique and valuable contributions, so neither can be benchmarked against simple comparables. The common version, the residual profit split, works in two steps. Suppose two affiliates jointly earn $100 million. First, each receives a market return for its routine functions: say $18 million and $12 million. Second, the residual, $100,000,000 − $18,000,000 − $12,000,000 = $70,000,000, is divided by the relative value of each side’s nonroutine contributions, for example 60:40 based on capitalized research spending: $42 million and $28 million. Totals: $18,000,000 + $42,000,000 = $60,000,000 and $12,000,000 + $28,000,000 = $40,000,000. The OECD rewrote its profit split guidance in 2018, and that text now sits in the 2022 Guidelines.
Facts (illustrative). A US parent buys components from its contract-manufacturing subsidiary in Country F, which taxes profit at 25%; the US rate is 21%. The subsidiary’s total costs are $200 million a year and the group’s combined pre-tax profit is $50 million. The subsidiary is the tested party; the profit level indicator is markup on total costs. Ten independent contract manufacturers earned: 3.1%, 4.2%, 4.8%, 5.5%, 6.0%, 6.4%, 7.1%, 7.8%, 8.5% and 9.6%. Assume Country F uses the interquartile range and adjusts to the median, as the US does.
The range. 25th percentile: 10 × 25% = 2.5, so the 3rd value, 4.8%. 75th percentile: 10 × 75% = 7.5, so the 8th value, 7.8%. Median: exactly half the results are at or below the 5th value, so average the 5th and 6th: (6.0% + 6.4%) ÷ 2 = 6.2%. Interquartile range: 4.8% to 7.8%.
Group tax. Every dollar of profit in Country F costs 25 cents instead of 21, so group tax = 25% × F profit + 21% × ($50 million − F profit).
| Scenario | Markup | F profit | Group tax if accepted | Group tax if challenged |
|---|---|---|---|---|
| A: below the range | 2.0% | $4.0m | $1.0m + $9.66m = $10.66m | F adjusts to 6.2%: +$8.4m taxed at 25% = +$2.1m; US already taxed it: $12.76m |
| B: bottom of the range | 4.8% | $9.6m | $2.4m + $8.484m = $10.884m | Inside the range, no adjustment: $10.884m |
| C: median | 6.2% | $12.4m | $3.1m + $7.896m = $10.996m | No adjustment: $10.996m |
The flip point. Against C, scenario A saves $10.996m − $10.66m = $336,000 a year if accepted (4% × $8.4 million), but costs $12.76m − $10.996m = $1,764,000 more if challenged and the US grants no offsetting relief (21% × $8.4 million taxed twice). A is better only if the probability of challenge, p, satisfies p × $1,764,000 < (1 − p) × $336,000, that is p < $336,000 ÷ ($336,000 + $1,764,000) = 16%. Interest and penalties pull that threshold lower still. A mutual agreement procedure (Section 4.4: Tax Treaties and Double Taxation Relief) can remove the double tax, but at the cost of years of negotiation. Scenario B keeps $112,000 a year of the saving with no adjustment, but sits on the edge: a cost overrun that pushes the actual markup to 4.7% takes it outside the range and back to the median. A reasonable policy targets the median and trues up before the books close.
The default (benchmark the simpler party with TNMM, accept the range) bends in these situations, all addressed in the OECD Transfer Pricing Guidelines 2022.
| Situation | What changes | Why |
|---|---|---|
| A routine entity reports losses year after year | Authorities adjust it back to a positive routine return | The Guidelines state that simple or low-risk functions are not expected to generate losses for a long period; independent firms would stop |
| Low value-adding services (payroll, IT help desk, bookkeeping) | A simplified cost plus 5% markup, with no benchmarking study | Guidelines paragraph 7.61 sets the 5% markup to cut compliance cost on low-risk flows |
| Commodity sales (metals, grains, oil) | CUP using the quoted exchange price on the pricing date, adjusted for differences | A public quoted price is the most direct comparable available |
| Both parties own unique intangibles | One-sided methods fail; use a profit split | No simple comparable exists for either side, so neither can be the tested party |
| Hard-to-value intangibles sold before their value is known | The tax authority may use actual results to revise the price after the fact | The 2018 guidance treats actual outcomes as presumptive evidence about the original price; one exemption applies when the gap stays within 20% of the compensation set at the time |
| Intercompany loans, guarantees and cash pools | Price the interest rate off the borrower’s own credit standing, including implicit group support | Chapter X (2020 guidance) treats financial transactions as a distinct class of transaction |
| Simple wholesale distributors in countries adopting Amount B | A preset return on sales from a pricing matrix replaces the benchmarking study | The OECD’s simplified approach (Section 4.5: Base Erosion and Profit Shifting (BEPS)) targets the most disputed low-risk cases |
Choose the method from the facts, then set the price at the middle of the range.
- If an identical product or license trades between unrelated parties on similar terms, use CUP (CUT for intangibles).
- If one party performs only routine functions, test that party with TNMM/CPM, using markup on costs for manufacturers and service providers and operating margin on sales for distributors.
- If both parties contribute unique intangibles or the operations are so integrated that neither can be tested alone, use a residual profit split.
- Target the median, not the edge of the range, and book a year-end true-up so the actual result lands there.
Assumptions and limits: the range is only as good as the comparables. If fewer than about six independent companies pass the screening, or their functions differ materially, widen the analysis or reconsider the method rather than trusting a thin range.
Setting prices in January and never checking them. Standard intercompany prices are fixed before the year starts; actual costs then drift. In the worked example, a subsidiary priced for a 6.2% markup whose costs overrun can finish the year at 2.0%, outside the range. The tax authority adjusts it back to the median, an $8.4 million adjustment, and unless the other country gives relief the group pays tax twice on it: 21% × $8,400,000 = $1,764,000 of double tax in a single year, before interest and penalties. A true-up invoice booked before the books close, moving the subsidiary back to 6.2%, costs nothing in tax.
Which transfer pricing method is used most often?
TNMM, called the comparable profits method in the US, is the most used. In the IRS’s report on advance pricing agreements executed in 2025, it was used for 86% of the tangible-property and intangible-property transactions covered. It dominates because operating margins of comparable companies are available from public databases, whereas exact comparable prices for the same product rarely exist.
What is the interquartile range in transfer pricing?
It is the middle half of the comparable companies’ results, from the 25th to the 75th percentile. US regulations accept a tested party’s result inside this range and ordinarily adjust a result outside it to the median. Other countries define the range differently, so the same data can pass in one country and fail in another (see the India Lens at the end of this Part).
When is the profit split method used?
Profit split is used when both parties to a transaction make unique and valuable contributions, such as two affiliates that each own key technology, or when operations are so integrated that neither side can be tested alone. Each party first earns a routine return; the leftover profit is then divided by the relative value of each side’s contributions.
CUP, resale price, cost plus, TNMM and profit split are the five OECD methods; TNMM (the US comparable profits method) dominates, and profit split applies when both sides contribute something unique. The worked example shows why pricing below the comparable range pays only if the chance of challenge is low (16% there), so target the median and true up before year-end.
Six questions on this chapter. Decide on your answer first, then click “Reveal Answer.”
1. Eight comparable distributors earned operating margins of 2.0%, 3.5%, 4.0%, 4.6%, 5.2%, 5.9%, 6.8% and 8.1%. A tested party earned 3.2%. Under the US interquartile range rule, to what margin is it ordinarily adjusted?
- 3.75%
- 4.9%
- 4.6%
- 6.35%
Reveal Answer
Answer: B. Range: 25th percentile = average of 2nd and 3rd values = 3.75%; 75th = average of 6th and 7th = 6.35%. 3.2% is outside, so it moves to the median, (4.6% + 5.2%) ÷ 2 = 4.9%.
2. Two affiliates jointly earn $80 million. Routine returns are $10 million for A and $6 million for B, and the residual is split 70:30 by the value of their unique contributions. What is A’s total profit?
- $56.0 million
- $44.8 million
- $54.8 million
- $50.0 million
Reveal Answer
Answer: C. Residual = $80m − $10m − $6m = $64m; A’s share 70% × $64m = $44.8m; total $10m + $44.8m = $54.8m.
3. Under TNMM, which entity should normally be chosen as the tested party?
- The entity that legally owns the group’s key patents
- Whichever entity sits in the highest-tax country
- The parent company, since it prepares the group accounts
- The simpler entity, such as a contract manufacturer
Reveal Answer
Answer: D. Reliable comparables exist for simple, routine functions, so the less complex party is tested and the other keeps the residual.
4. Pricing a subsidiary below the comparable range saves $200,000 a year if accepted but costs $1.8 million more if challenged and adjusted. Above what probability of challenge does the below-range policy stop paying?
- 10%
- 90%
- 20%
- 11.1%
Reveal Answer
Answer: A. Break-even p solves p × $1.8m = (1 − p) × $0.2m, so p = $0.2m ÷ ($0.2m + $1.8m) = 10%.
5. Worked problem: A group sells 5,000 units from one affiliate to another at cost plus 10%. Unit cost is $40. What is the price, and total revenue?
Reveal Answer
Answer: Price = $40 × 1.10 = $44.00. Revenue = 5,000 × $44 = $220,000.
6. Worked problem: The profit made by the seller on those units?
Reveal Answer
Answer: Profit = 5,000 × ($44 − $40) = $20,000, a markup on total costs of 10%.
- IRS Announcement 2026-8, Announcement and Report Concerning Advance Pricing Agreements (calendar 2025) — 110 APAs, 41.6-month median, 622 pending, India 35%, CPM/TNMM 86%