Cash Pooling: Physical vs Notional Structures, Controls and Accounting Handoffs

Search for cash pooling and you get a clean two-way split: physical pooling moves money, notional pooling does not. That is accurate, and it is roughly the first ten minutes of the decision. What decides whether a pool is worth running sits underneath it, and little of it is a banking question. A pool changes what each subsidiary holds, what the participants owe one another, what interest each books, and what a tax authority will expect you to prove about all of it.

The binding constraint is rarely the sweep. It is whether participants can lawfully lend to each other, whether the set-off the bank relies on is enforceable, whether the interest split survives an arm’s-length test, and whether anyone can reconstruct a given day’s position a year later.

The accounting, tax and transfer-pricing sections below were evidence-reviewed against OECD, IFRS Interpretations Committee, US Treasury, Federal Reserve and Chinese regulatory material current to 26 August 2026. They set out what those sources require and where the answer depends on the entities and jurisdictions involved.

The five structures and what each one actually moves

Cash pooling is a family of arrangements, not a single product. Each produces different balances, agreements and accounting entries even where the interest outcome looks similar.

The OECD guidance on financial transactions, now Chapter X of the Transfer Pricing Guidelines, treats physical and notional pooling as the two basic types and notes at paragraph 10.111 that groups combine them: several single-currency physical pools, say, with a notional pool over the header accounts. That overlay deserves separate treatment, because its guarantee requirements usually disappear when every header account belongs to one legal entity. Bank of America describes the same split from the provider side, and is explicit that notional pooling is offered only in certain locations.

What happens each day in a physical pool

Participating legal entities hold accounts at the pooling bank, each carrying a sweep instruction with a direction, a target, a frequency and a cut-off. At the cut-off the bank compares each balance against its target and moves the difference to or from the header account owned by the pool leader. Once every participant sits at target, the leader holds one residual position and either invests the surplus or draws on a facility to cover the shortfall.

Three details decide whether that is worth automating. If the pooling bank is not a direct member of the local clearing system, its investment deadline reflects an intermediary, which can pull a late-afternoon cut-off back to midday and cost a full day of interest. Movements inside one bank network are book transfers and typically cheap, while anything crossing a clearing system carries a fee and a settlement risk the design must absorb. And a movement that books today but values tomorrow leaves the interest benefit unclaimed and the intercompany balance misstated for a day. Intraday visibility decides how well any of this can be managed, which makes channel reliability part of the pool design rather than a separate technology question, and that decision is covered in our guide to bank connectivity.

Zero, target and threshold balancing

These three terms are used interchangeably in bank marketing, and they are not the same thing.

Zero balancing sets the target at nil. Every positive balance sweeps up, every negative balance is funded down, and the account ends the day flat. It captures the most liquidity and creates the largest intercompany position.

Target balancing leaves a stated amount behind, for an entity needing a buffer for payroll or tax, a bank requiring a compensating balance, or a regulator expecting the entity to hold its own funds. That retained balance is a deliberate cost, and should be sized from the entity’s own short-term outflows rather than set as a round number across the group. The 13-week cash flow forecast is where the number comes from.

Threshold or trigger balancing sweeps only once the balance passes a stated level, then usually moves the whole balance. Its purpose is cost control, and it means the account is not flat between sweeps, so both the interest calculation and the intercompany balance must handle days with no movement.

A structure can combine all three, with different rules by entity, currency and day of week. The weekend and holiday rule deserves attention, because a Friday sweep to zero leaves a participant with no buffer across a long weekend in a market whose clearing calendar differs from the header account’s.

Notional pooling and the bank’s own balance sheet

In a notional pool the participants keep their balances. The bank pays or charges interest on the net figure, either to a designated master account or across the participating accounts under a formula in the pooling agreement. No funds move, so there are no intercompany loans to book and the running cost is usually lower than an equivalent physical structure.

It is harder to obtain than that suggests, and the reason is the bank’s position rather than the customer’s. Deficit balances sit on the bank’s balance sheet as assets earning no normal lending return. To treat the pool as one exposure the bank needs an enforceable route to the credit balances, which is why the OECD notes at paragraph 10.113 that it will usually require cross-guarantees from participants to enable the right of set-off between accounts. Those guarantees are the real cost, and every participant gives them in favour of every other participant.

Appetite tightened once Basel III took effect as a Pillar 1 measure, because the Basel III exposure measure limits the netting a bank may recognise, and the liquidity coverage ratio attaches a runoff assumption to corporate deposits however the customer views the pool. Several banks repriced or withdrew multi-entity notional pooling as a result. That is bank economics rather than legal prohibition, and it varies by institution, so ask each bank in writing what it will offer, in which booking centre and on what guarantee terms.

The OECD adds a point groups frequently miss. At paragraphs 10.147 and 10.148 it observes that such cross-guarantees would not occur between independent parties: each guarantor covers every member without controlling membership or the amount guaranteed, and often cannot evaluate its real exposure. Where that holds, the guarantee may amount to no more than an acknowledgement of group support, no guarantee fee would be due, and support given on a default should be treated as a capital contribution.

Single-currency and multi-currency structures

Physical pooling is a single-currency technique. Funds can only sweep between accounts in the same currency, so a group operating in five currencies runs five pools, each with its own header account, residual position and funding decision. Cross-border works within a currency where funds may move freely; the currency boundary is the hard one.

Notional pooling can cross currencies, which is its main structural advantage and also where the cost hides. Balances must be brought to a common base currency before they can be offset, through either a short-dated swap or a notional conversion with the currency risk recovered in adjusted rates on each leg. Neither is normally quoted as an explicit fee, so a pool that looks cheap on its tariff can carry a conversion spread larger than the benefit it produces. Compare all-in yield against running the currencies separately.

Interest optimisation sits between the extremes. It converts end-of-day balances notionally across countries and currencies and pays an enhanced rate on the combined figure with no movement of funds; its real attraction is that it can include balances in markets where funds cannot leave, though the rate is usually modest. Where pooled balances sit outside an entity’s functional currency, the consequences are covered in our guide to remeasurement and translation.

Intercompany positions and interest allocation

Physical pooling converts a bank balance into a related-party balance. Each sweep moves funds between two legal entities, and unless every account belongs to the same entity it creates a loan. Run daily, that is a revolving intercompany position needing a rate, a day-count basis, an accrual, a settlement cadence and an agreement behind it. The common failure is not mispricing but silence: interest is never accrued at participant level, the balance grows for a year, and nobody can say what the entity owes or on what terms. Agreeing, matching and settling those balances belongs to the same discipline as intercompany netting and settlement.

How the benefit is split among participants is a pricing decision, not an administrative one. The benefit itself is arithmetic: netting a debit position against a credit position removes the bank’s spread on the offset amount. The OECD approach at paragraphs 10.143 to 10.146 is to fix the leader’s reward first, then allocate what remains through arm’s-length rates on each participant’s debit and credit position. Two constraints follow. Every participant must be better off than its next best option, which paragraph 10.146 broadens to benefits other than rate, such as access to a permanent funding source or to liquidity otherwise unavailable. And the split cannot be arbitrary, because one set to move income between jurisdictions is exactly what tax authorities look for.

The leader’s own reward follows its function. Paragraph 10.130 states that a leader generally performs no more than a co-ordination or agency function, the header account acting as a centralised point for book entries meeting predetermined target balances, and that its remuneration as a service provider will be similarly limited. Keeping part or all of the spread requires the profile in the OECD’s second example: an entity that sets intra-group rates, bears the difference against external rates, carries credit, liquidity and currency risk, and both controls those risks and can bear them.

Tenor matters too. The OECD frames pooling as short-term liquidity management at paragraph 10.110, and at paragraphs 10.122 and 10.123 asks whether a position that has stopped behaving like a short-term balance should be delineated as something else, such as a term deposit or loan. Its practical test is whether the same pattern recurs year after year. A participant that has been a net borrower of a similar amount for three consecutive years is holding term funding through a pooling wrapper, and pricing it overnight is hard to defend.

What the pool does to your accounts

Two accounting questions arise, with different answers at different levels of the group. The first concerns a participant’s own financial statements. Deloitte’s roadmap on centralized cash management arrangements states that funds a subsidiary deposits into a parent’s cash account should not be classified as cash or a cash equivalent in the subsidiary’s separate financial statements where the subsidiary does not hold legal title to them. The deposit is not a demand deposit at a bank and does not meet the cash equivalent definition, so it is an intercompany receivable. That flows into the cash flow statement: increases in receivables from affiliates are investing activities and increases in payables to affiliates financing activities, generally gross, with net presentation acceptable where turnover is quick, amounts large and maturities short, or amounts due on demand. A subsidiary reporting its pool contribution as cash is overstating liquidity it does not control.

The second concerns whether debit and credit positions can be presented net. Under US GAAP, ASC 210-20-45-1 permits offset only where all four conditions hold: each of two parties owes the other determinable amounts, the reporting party has the right to set off, the reporting party intends to set off, and the right is enforceable at law. IFRS reaches the same structure through IAS 32.42, and the Interpretations Committee agenda decision of March 2016 addresses cash pooling directly. The submitter’s group had an enforceable right of set-off at the reporting date and made regular physical transfers into a netting account, though not on the reporting date itself. The Committee concluded that a right is not enough: where a group expects its subsidiaries to use their individual accounts in the normal course before the next net settlement date, it would not be appropriate to assert an intention to settle the entire period-end balances net, because that presentation would not reflect the expected amounts and timings of future cash flows.

That conclusion is narrower than it is often reported to be: the Committee noted expressly that expectations differ across arrangements and that judgement is required. The operative point is that the offsetting analysis turns on what participants are expected to do with their accounts, an operating fact the pool design controls, and not on the set-off clause alone.

Tax, transfer pricing and the documentation set

Everything in this section depends on which entities participate and where they are resident. Legal availability, how interest is characterised and taxed, whether withholding applies, whether the pricing meets local transfer-pricing standards and what the accounting conclusion is are all determined by the specific entities and jurisdictions involved, and require review by tax, legal and accounting specialists engaged on the actual structure. What follows identifies what those specialists will be applying, not a conclusion for any group.

Start with characterisation, because it drives most of the rest. If the header account owner acts as principal, the interest is intercompany interest and exposed to withholding. If it acts as agent for the participants, the interest may retain the character of bank interest, treated differently in some jurisdictions. That distinction lives in the agency agreement, a drafting decision made before the pool opens.

Withholding is where cross-border pools most often stop being economic. A United States participant paying interest to a foreign affiliate cannot generally rely on the portfolio interest exemption, because section 881(c)(3)(B) excludes interest received by a ten per cent shareholder and interest received by a controlled foreign corporation from a related person. Without treaty relief the default is a thirty per cent charge on the gross interest, which usually exceeds the pooling benefit outright. Relief is often available but depends on residence, limitation-on-benefits provisions and beneficial ownership, none of which follows from incorporating an entity in a treaty jurisdiction.

Two further United States points affect group structures. Related-party pricing has a domestic reference point: Treasury Regulation section 1.482-2 provides a safe haven for certain dollar loans at between one hundred and one hundred and thirty per cent of the applicable federal rate for the relevant term. Separately, section 956 can produce a deemed dividend where a controlled foreign corporation is treated as investing in United States property, which is why combining United States and non-United States pools, or taking collateral support from such a corporation, needs specific advice. Final regulations issued in May 2019 reduced that inclusion for corporate United States shareholders to the extent a section 245A deduction would have been available on an equivalent distribution, so pre-2018 commentary is unreliable.

Documentation is not paperwork produced after the fact. A physical structure normally needs a pooling agreement with the bank, an accession agreement per entity, an intercompany loan framework stating rate basis, day count, accrual and settlement, board resolutions recording corporate benefit for each participant, account mandates and know-your-customer records. A notional structure replaces the loan framework with cross-guarantee and set-off deeds, a heavier legal commitment rather than a lighter one. The OECD notes at paragraph 10.124 that participants may be resident across many jurisdictions, making the position hard for any single tax authority to verify, and asks groups to describe the structure and the returns to leader and members in their transfer-pricing documentation.

Where each structure is actually available

Availability is a live question rather than a settled one. The positions below are dated, and should be confirmed with local counsel and with the banks concerned before a structure is designed around them.

The Association for Financial Professionals, writing on concentrating cash across borders, names the Brazilian real and Indian rupee among currencies that cannot be freely converted, and notes that banks may decline notional arrangements without compensation for the regulatory capital they must hold.

The United States row needs care, because the common claim that notional pooling is illegal there is repeated with several incompatible explanations and no clean statutory citation. What can be said is that United States banks do not generally offer multi-entity notional pooling with interest offset on domestic accounts, that cross-border pools are typically booked elsewhere, and that the constraints practitioners cite involve deposit-insurance treatment, regulatory reporting and the enforceability of cross-entity set-off. The question was still being worked through as recently as 2022, when the Bank Policy Institute wrote to the banking agencies about how notional pool balances should be reported on Call Report Schedule RC-O. Treat it as a bank-by-bank question to put in writing, not a settled rule. Physical concentration carries its own consequence: FDIC deposit insurance covers $250,000 per depositor, per insured bank, per ownership category, so concentrating balances at one bank removes a diversification the unpooled structure had.

Two United States constraints that older pooling literature still cites have in fact been removed, and repeating them leads to poor design. The prohibition on paying interest on demand deposits was repealed by section 627 of the Dodd-Frank Act from 21 July 2011, and the Federal Reserve repealed Regulation Q accordingly. The six-transfer monthly limit on savings deposits, which shaped sweep design for decades, was deleted by an interim final rule amending Regulation D in April 2020 after reserve requirements were set to zero.

China is moving in the opposite direction. The People’s Bank of China and the State Administration of Foreign Exchange upgraded their integrated local and foreign currency pooling policy across ten pilot regions in December 2024, permitting cross-currency lending among onshore participants for current-account cross-border payments and letting groups set their own external debt and overseas lending proportions. Draft measures published in April and July 2025 extend the regime nationwide, so any China design should be checked against the version in force for the specific region and entity.

Controls, ownership and the exception path

A pool concentrates cash, and anything that concentrates cash concentrates the consequences of a control failure. The parameters are the sensitive asset: whoever can change a target balance, a sweep direction or a participant list can move money without initiating a payment. The appetite and concentration limits that justify those parameters belong in the treasury risk management framework before configuration begins.

The exception path matters more than the happy path, because the happy path runs unattended. Four failures recur, each needing a named owner and a decided response before it happens rather than during it. A sweep that does not execute leaves a participant unfunded and possibly overdrawn against a facility it does not have, so the control is a next-morning exception report driven by expected against actual movement, not by the absence of an alert. A missed cut-off leaves the balance in place for a day, tolerable financially but only if reflected in the interest calculation rather than silently ignored. A suspended participant has to leave the structure without stranding an unsettled balance, so the removal procedure is drafted at implementation and not improvised. And a header account closing in an unexpected position points at a failed sweep or an unforecast flow, and should route to a person rather than a dashboard.

These controls must survive automation rather than be replaced by it, the subject of our guide to treasury automation. Where the structure outgrows spreadsheet administration, the capabilities to test in a platform are set out in our guide to the treasury management system category.

Reconciliation and the daily evidence

Reconciling a pooled structure is not the same as reconciling a bank account, because two ledgers must agree with one statement set: bank statements against the participant accounts, and the pool leader’s intercompany ledger against the same movements from the other direction.

Three artefacts do the work. The end-of-day statement, an ISO 20022 camt.053 message or its SWIFT MT940 or BAI2 equivalent, is the only one carrying booked entries with final balances, so it is what the ledger reconciles against. Intraday reporting, camt.052 or MT942, shows movements as they book and is what treasury uses to manage the position during the day: a management tool, not a reconciliation source. The bank’s interest statement is the third, and should be recalculated rather than accepted, because the interest split is a pricing position the group has to defend.

Four daily checks are enough for most structures. Confirm every expected sweep executed and every executed sweep was expected. Confirm each participant account closed at its stated target, and investigate any that did not. Confirm the sum of participant movements equals the header account movement for the day. Confirm each movement created a matching intercompany entry with the same value date, because a booking-date entry against a value-dated movement produces a difference that compounds quietly. Where the position feeding these checks is assembled automatically, the data chain is covered in our guide to automated cash positioning.

Period end needs one more step. The cut-off used for the last sweep of the period must be the cut-off used for the accrual, and the offsetting conclusion has to be applied to the balances as they stand at the reporting date, not as they stood on the last day a full sweep ran.

Choose the structure, then prove it works

The decision resolves quickly once the constraints are stated in the right order. If participants cannot lawfully lend to each other, or intercompany interest attracts withholding exceeding the benefit, the physical structure is out regardless of its efficiency. If cross-guarantees cannot be given, because a board cannot establish corporate benefit or a lender’s covenants prohibit them, notional pooling is out. If positions are structurally long-term, the answer is a term intercompany loan and not a pool at all. Only once those gates are passed does the interest arithmetic decide anything.

That arithmetic is simple and worth doing before any bank conversation: the recurring benefit is the offset amount multiplied by the difference between the debit and credit rates. As a Finance Circuit illustration rather than market data, a group with $40 million of participant debit positions offset against credit positions, paying 6.2 per cent on debit balances and earning 3.9 per cent on credit balances, captures a 230 basis point spread on the offset amount, or roughly $920,000 a year. Against that sit the four cost categories in Treasury Alliance Group’s treasurer’s guide to pooling: implementation, maintenance, per-transfer cost and opportunity cost. The last is never quoted, and covers rates below what the group could earn directly, conversion spreads on multi-currency structures and any earlier cut-off imposed by the pooling bank. On a marginal structure it decides the answer.

Before go-live, test the structure rather than the diagram. Remove a participant and confirm the remaining instructions still resolve. Fail a sweep deliberately and confirm the exception reaches a person. Insert a cross-currency balance and confirm it is rejected rather than converted silently. Recalculate one month of allocated interest independently of the bank. Where that needs system support, the selection criteria are in our guide to liquidity management software, and where the pool sits alongside the ledger and the bank feeds, our finance systems integration map covers the ownership boundaries. Where pooling materially changes bank, lender, maturity or entity concentration, the resulting limits and stress treatment belong in the liquidity risk management layer.

One boundary is worth naming. A pool leader that sets rates, bears the difference against external funding, carries credit and liquidity risk and decides how the group is funded has stopped being a coordinator and is operating as an internal bank, with a different pricing analysis, control set and documentation. That is a separate operating model and a separate decision. This article is current through 26 August 2026.