The facts
A… Lda (the Taxpayer) is a Portuguese limited liability company incorporated in 1995, manufacturing industrial textile belts for mining and heavy transport and specialist transmission belts for the automotive sector. It is 74.5 per cent owned by B… and 25.5 per cent owned by C…, both members of the D… Group, whose ultimate parent E… is German.
The Taxpayer participated in two cash pooling arrangements, both zero-balance pools. Under the euro pool with E…, it paid 4 per cent on debit balances and received 2 per cent on credit balances. Under the US dollar pool with C…, the rates were 4.5 per cent and 2 per cent respectively. The Taxpayer also held a €12,000,000 long-term shareholder loan from C…, originally at 4.5 per cent and reduced to 4.0 per cent in 2017, which rate applied throughout 2019 and 2020.
In its transfer pricing documentation for those years, the Taxpayer included — as contextual background only, a detail that would come to define the Tribunal’s reasoning — data from the Banco de Portugal’s monthly statistical bulletin: weighted averages of interest rates on new credit operations above €1 million extended by Portuguese financial institutions to euro-area residents.
In 2023, the Serviços de Inspeção Tributária (SIT) conducted partial-scope IRC inspections for 2019 and 2020. On the cash pooling, the SIT rejected the spread between debit and credit rates, arguing that because E… and C… are not financial institutions the rates should be symmetrical. On the shareholder loan, the SIT rejected the 4.0 per cent rate as above market. In both cases the authority used the Banco de Portugal averages — drawn from the Taxpayer’s own file — as the CUP benchmark. Total IRC and withholding tax corrections across both years amounted to €158,063.83, paid by the Taxpayer in August 2023. An administrative claim (reclamação graciosa) was rejected in October 2024. Arbitral proceedings commenced in January 2025.
The litigation
CAAD Tribunal Arbitral Colectivo · 22 August 2025 · taxpayer prevails
The Tribunal identified two issues: whether the arrangements complied with the arm’s length principle under Article 63 CIRC; and whether the authority’s CUP methodology met the comparability standard required by that provision and Portaria n.º 1446-C/2001.
On the cash pool spread, the Tribunal rejected the authority’s symmetry argument. Debit and credit positions in a zero-balance pool are not the same transaction. The pool manager performs treasury coordination and management functions that — where no separate fee is charged — are legitimately reflected in the spread between debit and credit rates. Requiring equal rates in both directions ignores the economic content of the pool manager role. The authority had also argued that the spread was commercially irrational, citing a decline in equity from €8.85 million in 2018 to €3.26 million in 2020. The Tribunal declined to substitute the authority’s commercial judgment for that of the Taxpayer’s management; equity erosion across two years does not establish a TP violation. The Tribunal followed CAAD Processos 253/2019-T and 140/2023-T, which had reached identical conclusions on materially similar facts.
On the CUP methodology, the Banco de Portugal averages do not satisfy the highest comparability standard mandated by Article 63(2) CIRC and Article 6(1) of the Portaria. The benchmark covers every new euro-denominated credit operation above €1 million made by Portuguese banks to euro-area companies across a full calendar year. It inevitably includes transactions with radically different maturities, security structures, covenants, borrower profiles, and commercial purposes. It cannot serve as a valid CUP comparator for specific intercompany financing arrangements. The same analysis disposed of the shareholder loan correction, following CAAD 253/2019-T, 140/2023-T, and 687/2017-T.
The withholding tax assessments were dependent on the underlying TP corrections and fell away on annulment. Compensatory interest was awarded from the date of rejection of the administrative claim (October 2024), following the Supremo Tribunal Administrativo uniformisation of jurisprudence of 29 June 2022 (process 093/21.7BALSB): the error becomes attributable to the authority at the point it rejects the administrative challenge, not at the date of original assessment.
ajiho commentary
The authority used the taxpayer’s own background data as its primary benchmark
The Banco de Portugal averages the authority used to challenge the Taxpayer came from the Taxpayer’s own transfer pricing documentation, where they had been included as contextual market data. The authority treated background context as a primary CUP comparator. The Tribunal was direct: these are different things, and the obligation to conduct a proper comparability analysis cannot be discharged by repurposing materials the Taxpayer assembled for a different purpose.
The practical lesson is clear. Including broad statistical data for context is legitimate and sometimes necessary. But if that data is capable of being misread as a benchmark, the file must be explicit about what it is and what it is not. A scope note or methodology statement that distinguishes background market data from the primary comparability analysis is not optional — this case is the demonstration of what happens without one.
Four decisions, four rejections, same methodology
The line runs 687/2017-T, 253/2019-T, 140/2023-T, and now 68/2025-T. The authority has deployed the Banco de Portugal weighted averages as a CUP benchmark in each case, and the Tribunal has rejected it in each case on the same ground: the sample is too broad and too heterogeneous to satisfy the highest comparability standard the CUP method demands. What is notable is that the authority continued to deploy this methodology after 140/2023-T, a decision whose rejection of the benchmark was unambiguous. Four consecutive rejections on the same ground raise a legitimate question about whether continued reliance is a defensible audit approach. A fifth case on this point would be very hard to defend.
The spread is the price of functions the authority never charged for separately
The authority’s symmetry argument — that equal rates must apply because the pool manager is not a bank — is conceptually wrong. The arm’s length question is not whether the pool manager resembles a bank. It is whether independent parties in the same position would agree to different rates for lending and borrowing within the pool. Where the pool manager provides treasury management, liquidity coordination, and risk monitoring functions without a separate fee, pricing those functions through the rate differential is a recognised and commercially rational approach.
Note what the Tribunal did not resolve: whether the specific spread in this case was within the arm’s length range. The methodology failed before that question was reached, so the quantum was never tested. A taxpayer who prevails on methodology alone has won the battle without necessarily having demonstrated that their specific pricing is arm’s length. One further dimension for Iberian groups: Portugal and Spain have reached opposite conclusions on symmetry. The CAAD has now twice endorsed the spread as legitimate; Spain’s Tribunal Supremo has ruled symmetry is required. Groups with pools in both jurisdictions are caught between irreconcilable national positions on the same structural question.
What this means for your business
If you have an open Portuguese TP correction on cash pooling or intercompany loans built on Banco de Portugal averages, four arbitral decisions have now rejected this benchmark on the same ground. The grounds for challenge are established and consistent, and the RJAT arbitral process has been used successfully in materially identical circumstances.
If you are documenting cash pool arrangements with a debtor/creditor spread, the spread needs to do two things: be clearly linked to the pool manager’s treasury management functions and the absence of a separate fee for those functions; and be supported by specific arm’s length evidence for the quantum of the spread itself. A spread is permissible in principle, but that principle does not tell you whether your specific spread is within the arm’s length range.
If you include market statistics in your TP documentation as background context, be explicit in writing about what that data is. Background statistics are context, not a benchmark. A scope note or documentary flag that clearly distinguishes background market data from your primary comparability analysis is essential — this case shows what can happen when an authority treats the former as the latter.
Case reference
A… Lda v Autoridade Tributária e Aduaneira · CAAD — Centro de Arbitragem Administrativa e Fiscal, Tribunal Arbitral Colectivo · Processo n.º 68/2025-T · 22 August 2025
Arbitrators: José Poças Falcão (President), José Nunes Barata (Rapporteur), José Joaquim Sampaio e Nora. Cases cited: CAAD 687/2017-T, 253/2019-T, 140/2023-T; Tribunal Central Administrativo Sul 18/18.7BCLSB (13 December 2019); Supremo Tribunal Administrativo 093/21.7BALSB (29 June 2022). Drafted from the primary CAAD decision.