Which side of the transaction should be the tested party?
The side whose functions are simpler, whose risks are fewer, and for which independent companies doing something similar can actually be found. That is one test rather than three: complexity is what makes comparables scarce, so the less complex party is the one whose margin can be measured against outside data with a straight face. Size is irrelevant and so is which entity the group happens to think of as the important one. Once the choice is made, it has to be supported by a functional and risk analysis describing what each party does, decides and funds, because the selection is the first thing an examiner questions.
Can the parent company be the tested party?
Yes, and treating that as impossible is a common mistake. Nothing about ownership makes an entity complex. A parent that holds shares and provides routine administrative support while the operating subsidiary develops the technology, carries the market risk and funds the launches is the simpler party on the facts, and it is the one whose return can be measured. What decides the question is conduct, not the organisation chart. The practical difficulty is usually evidential rather than conceptual: the parent's accounts often mix the tested support activity with holding-company items, so a segmented result has to be built before any comparison can be made.
What if neither party is clearly the simpler one?
Then say so in the documentation rather than picking one and hoping. Where both sides contribute things for which no outside price exists, no single margin can stand for the whole arrangement, and a method that divides the combined result by reference to each side's contributions is the honest answer. The important discipline is that this conclusion is reached through evidence and recorded, because the alternative pattern, selecting the convenient party and describing the other as a mere risk-bearer, tends to survive exactly until the counterpart country reads the same facts from its own side and reaches the opposite conclusion.
Do I need separate accounts for the tested party?
Almost always, because the entity and the tested activity are rarely the same thing. If a company sells to related parties and to outside customers, or runs a second line of business, its entity-level result blends activities the comparable companies do not perform, and the comparison then measures the wrong business. What is needed is a segmented profit and loss for the tested activity alone, built from the ledger, with a stated basis for every allocated cost and every shared overhead. Keep the allocation basis constant from year to year and record which allocations required judgement, so the same segmentation can be reproduced and defended later.
Can the tested party change from one year to the next?
It can, but only because the facts changed, and the file has to show which facts. A restructuring that moves risk control or intangible ownership across the border genuinely changes which party is the simpler one, and the analysis should follow. What does not justify a change is a poor result: switching sides because the other entity's figures compare better is visible immediately, since the functional description will not have moved with it. Where a change is real, document the date it took effect, the decisions and funding that moved, and expect the counterpart authority to look closely at the year in which the switch happened.
Is the tested party the same as the entity being audited?
No, and the confusion causes real trouble. The tested party is an analytical choice about whose margin the documentation measures, made when the method is designed. Which entity receives a query is a matter for each tax authority, and an authority will happily examine its own resident taxpayer while the documentation measures the affiliate on the other side of the border. That situation is normal and workable, provided the local entity can produce the whole analysis, including the counterpart's segmented figures and the search behind the range. Where it cannot, the examination stalls on evidence rather than on the merits of the pricing.
Which countries have a tax treaty with the United States?
Around sixty, including Canada, the United Kingdom, India, Australia and most of western Europe — but the list matters less than the terms, because each treaty caps rates and allocates income differently. Two countries with treaties can produce opposite answers on the same pension or the same royalty. What decides your position is the specific article covering your income type. See our country guides.
Do NRIs pay tax on money sent to India?
Sending your own funds to India is a transfer of capital, not income, so the remittance itself is not taxed. What is taxable is income the money then earns in India — interest, rent, capital gains — under the rules for the account type it sits in. Sending money out of India is the direction that needs certification before the bank will act. See NRE, NRO and FCNR accounts.
