Build a Benchmark
Run a benchmark study the way a transfer pricing team does: search, screen, find the range, then test the Canada Distributor.
First, a guess
The Canada Distributor buys finished products from the US Entrepreneur and resells them in Canada. It takes little risk and aims for a 3% Operating marginOperating profit divided by sales. It is the most common profit level indicator for distributors.Open in the Glossary. To show that is arm's length, the team compares it with independent distributors that do similar work: a Benchmark / comparablesA benchmark is a study of independent companies or transactions (the comparables) that are similar enough to the tested party. Their results form the arm's length range.Open in the Glossary.
Run the study
Four steps, in order. The Distributor is the Tested partyThe company in a transaction whose results are compared with comparables. It is usually the simpler, less complex party, such as a routine distributor or manufacturer.Open in the Glossary: its result is the one being checked.
Search a database
The Distributor buys finished goods and resells them. So the team searches a company database for wholesale distributors in North America with public accounts. Each company's result is its operating margin averaged over 2023 to 2025.
- Industry
- Wholesale distribution of goods
- Region
- US and Canada
- Years
- 2023 to 2025
- Result
- Operating margin, 3-year average
Screen each company
Read each business description and decide: accept, or reject with a reason. Keep the screening criteria open while you work. The margin is shown, but it never decides the call.
Run the search first. Screening starts from its results.
Find the range
Finish screening first. The accepted companies make the range.
Test the Canada Distributor
Finish screening first. The Distributor is tested against the accepted companies.
Read the company, then count the middle
A benchmark is only as good as its companies. Screening asks one question about each candidate: does it do the same kind of work, with the same kind of risks, as the tested party? That is why you read the description, not the margin. Pinecrest (3.2%) and Ostrava (2.6%) had normal margins and still went out. Laurentide (1.2%) and Bluewater (5.6%) had the lowest and highest margins and stayed. Dropping companies because their results look high or low is cherry-picking, and tax auditors look for it.
US rules list what makes companies comparable: their functions, contract terms, risks, economic conditions, and the goods or services involved. Under the Comparable profits method (CPM)A US method that tests a company's operating profit against the operating profit of comparable companies, using a profit level indicator such as operating margin (Treas. Reg. § 1.482-5). It is close to the OECD's TNMM.Open in the Glossary, similar functions and risks matter more than an exact product match. In practice, one bad year is not a reason to reject a company; a business that keeps losing money usually does not match a low-risk distributor.
Then sort the accepted results and keep the middle half, from the 25th percentile (Q1) to the 75th (Q3). The result in the middle is the median. Because the range counts places, not distances, one far-off company barely moves it: adding a 25% company moved Q3 from 4.0% to 4.4% under the US rule, not toward 25%.
The IRS uses the Interquartile range (IQR)The middle 50% of comparable results, from the 25th to the 75th percentile. The US rule (Treas. Reg. § 1.482-1(e)(2)(iii)(C)) picks actual comparable results or averages two of them, while Excel's PERCENTILE.INC interpolates between values, so the two can give slightly different ranges.Open in the Glossary when the comparables are good but not perfect, which is almost always. If the tested party lands outside it, the IRS ordinarily adjusts the result to the Median adjustmentIf a US taxpayer's result falls outside the interquartile range, the IRS will ordinarily adjust it to the median of the comparables (Treas. Reg. § 1.482-1(e)(3)).Open in the Glossary.
Sources: Treas. Reg. § 1.482-1(d), comparability; § 1.482-1(e), the range and the median
Check yourself
Three quick questions. Each answer explains itself, right or wrong.
0 of 3 answered.
Try it for real
- Screen real companies in the ScreenerSearch SEC filers, read each business description, and accept or reject with a matrix like the one above.
- Build a real range in BenchmarksGet Q1, the median and Q3 from real companies, the US way or the Excel way.
- Ask the AI AssistantTry: should a distributor with one loss year stay in my comparable set?