Profit Split Method: When One-Sided Methods Break Down

Every one-sided method — CUP, RPM, cost plus, TNMM, CPM — works by pricing one party’s contribution against a market reference and leaving the rest of the profit where it falls. That works when one party’s contribution is routine: it has a market price or a market margin, so pricing it against comparables resolves the transaction.

It breaks down when both parties contribute something genuinely non-routine — valuable intangibles, unique functions, significant risk-taking — or when the transactions are so integrated that they cannot be priced separately. That is the profit split domain.

The two types of split

	Conventional split	Residual split
Starting point	Relative contributions of the parties’ functions, assets and risks	First give each party a routine return, then split what remains
Intangibles	No unique, valuable intangible is attributed to either party as the driver	Residual profit is allocated in proportion to the relative value of the unique intangibles/entrepreneurial functions
Data needed	Both parties’ financials + contribution analysis	Same, plus a defensible valuation of the residual drivers
Typical use	Integrated operations, shared platforms, co-development with comparable market evidence	One party (or both) holds HTVI or entrepreneurial value the market does not price

The conventional split compares the relative economic contributions of the two parties (or both) and divides the combined profit accordingly. The residual split does a two-step: pay each side its routine return (often cost-plus on the routine functions), then allocate the leftover profit based on the relative value of what is not routine — the intangibles, the entrepreneurship, the control of risk.

Residual splits and hard-to-value intangibles

The residual split is the standard answer to the HTVI problem — intangibles that are valuable but for which no market price exists (early-stage technology, proprietary algorithms, the potential of a new platform). There is no CUP for “the value of this engine”, so a one-sided method that prices the routine side against comparables silently attributes all the residual value to the side that owns it — which may be wrong in either direction.

The residual split forces the attribution to be explicit: the file must say what the routine return is (with the benchmark behind it), what the residual is, and what relative-value argument justifies the split ratio. That explicit attribution is both the method’s strength and its exposure — the allocation key becomes the item the authority examines.

Worked example: co-development software

Party A (India) develops and maintains a software platform under contract; Party B (US) owns the platform IP, sets the product direction and sells subscriptions to third parties. FY 2025-26:

Item	Party A	Party B
Revenue	400 cr (fees from B)	5,000 cr (subscriptions)
Operating cost	320 cr	3,600 cr
Operating profit	80 cr (20% of cost)	1,400 cr

A one-sided TNMM on Party A gives A its routine cost-plus return — defensible, and it leaves the entire 1,400 cr residual with B. If A’s development work is a key value driver (B’s sales story is built on the platform A engineers), the file argues the residual split:

1. Routine returns first. A is paid its benchmarked cost-plus (say the 25th–75th percentile range for comparable Indian developers — which may include the booked 20%). B’s routine distribution function is paid its limited-risk return.
2. Residual. What remains after the routine returns — the platform’s entrepreneurial profit — is the residual pool.
3. Allocation key. DEMPE analysis: A performs Development and significant Enhancement under B’s direction; B performs the commercial exploitation and bears the market risk. The file proposes a split (for example 30:70) with the contribution analysis attached.

The audit question is not the arithmetic. It is: who performed which DEMPE function, who controlled the risk, and what evidence prices the key contributions relative to each other — see DEMPE and routine vs entrepreneurial.

The comparability burden

Profit split is two-sided, and that changes the evidence standard:

• Both parties’ data must be available and reliable — the tested party’s accounts alone are not enough.
• There is no pool. You cannot benchmark “the split ratio” against a database of similar splits; the comparability argument is built from the parties’ own functions, assets, risks and market evidence.
• Integration matters. If the transactions cannot be separated at all (shared inputs, pooled working capital, cross-licensed IP), the split is usually the only method that survives — but the file must show the integration, not just assert it.

TPO practice in India

Indian authorities are cautious with the profit split:

• TPOs frequently reject the split and substitute TNMM on the Indian party on the reasoning that the Indian side’s functions are in fact routine — which, if true, defeats the split’s own premise. The taxpayer’s answer is the DEMPE and risk-control record: contracts, board minutes, decision rights, who funded the development, who bore the loss years.
• Where the Indian entity genuinely performs key development (a true co-development, not a contract), the split argument is strongest in appeals (AAR/ITAT), where the functional record is examined properly.
• The documentation must contain the entire pool of both parties’ financials and the allocation analysis — a split argued on one party’s books is a split the TPO can disregard.

Documentation checklist

• The integration analysis: why no one-sided method prices the transaction.
• The routine-return benchmarks (with the underlying Accept-Reject matrices) for the first step of a residual split.
• The DEMPE allocation with primary evidence for each function.
• The allocation key, its sensitivity, and the fallback position if the key is rejected.
• The year-by-year consistency: the same key logic applied across years, with the variance explained when outcomes move.

See also

• Profit Split vs TNMM
• Tested party selection
• Transfer pricing risks