Intercompany reconciliation: why the two sides disagree

Most intercompany differences are not errors. Five structural reasons two ledgers disagree, six dimensions that decide the automation, and the policy question a matcher cannot answer.

By the Foxnut team · Updated 11 August 2026

Why two ledgers disagree about the same transaction

Intercompany reconciliation is the exercise of agreeing what one entity in a group says it is owed against what the counterpart entity says it owes, before both sides are removed from the consolidated accounts. The removal itself is not negotiable. IFRS 10 requires consolidated financial statements to “eliminate in full intragroup assets and liabilities, equity, income, expenses and cash flows relating to transactions between entities of the group”, so a pair that does not agree cannot be eliminated in full and the difference has to be put somewhere. What makes this process unlike every other reconciliation in finance is the reason the pairs disagree. A bank reconciliation has one right answer sitting on a statement. An intercompany pair often has two right answers, because the accounting standards and the tax rules require the two sides to look different. The three-month gap a subsidiary is allowed between its reporting date and the group’s, the exchange difference the standard forbids anyone to eliminate away, the recharge one entity books as revenue and the other as a cost recovery, the year-end transfer pricing adjustment reported on a return with no book entry behind it: none of these is an error, and none of them is fixed by matching harder. The practical question is therefore not how to agree the balances faster. It is how many of the differences a system can be expected to close on its own, and who resolves the rest. That second half is why how a reconciliation system reaches the people who run it is worth settling before the software question, because the residual population is the whole of the work.

Five reasons a matched pair still fails to match

The distinction that matters is not big difference against small difference. It is whether the difference is a mistake or a requirement.

Why the pair differs	What the two ledgers show	Can a matcher settle it
Cut-off	One side posted the transaction in the period, the other in the next. IFRS 10 permits up to three months between a subsidiary’s reporting date and the group’s, provided the gap is the same from period to period	It can pair the two entries across the boundary. It cannot decide which period is right, because that is the group’s cut-off policy
Currency	The same balance carried in two functional currencies. Where an intragroup monetary item is denominated in one side’s functional currency, the exchange difference arises in the other side’s books and not in both	No, and it must not. The standard says an intragroup monetary item cannot be eliminated against its counterpart without showing the currency effect in the consolidated statements
Accounting policy and classification	One entity treats a management recharge as revenue, the other as a recovery of cost. IFRS 10 requires uniform policies for the consolidated statements, not identical treatment inside every entity	No. Somebody has to make a policy call and apply it in both ledgers, which is a decision rather than a match
Transfer pricing adjustments	A year-end true-up posted on one side only, or no book entry at all: a controlled taxpayer may report a controlled transaction on a timely filed United States return at a price different from the one actually charged	No. The difference is between a tax basis and a book basis, and it is deliberate
Non-submission and disputes	One counterparty sent incomplete information, made an entry error, or holds a legitimate difference of opinion about how the transaction should be treated	Partly. It surfaces the open item on day one instead of day ten. It cannot make the other entity answer

Read down the third column and the shape of the residual is already visible. One of the five is a matching problem. Two are policy decisions. One is a filing convention. One is a dependency on another team.

What the decision turns on

Six structural dimensions decide whether a process is worth automating. Intercompany reconciliation reads unusually on two of them: its exceptions are manufactured by the group’s own legal and currency structure rather than by data quality, and its regulatory exposure runs in two directions at once, because the same intercompany number is filed to the market on one basis and to tax authorities on another.

Intercompany reconciliation: process profile
Dimension	What it reads on intercompany reconciliation	Source
Exception variance	The standard case is a pair posted in the same period, in the same currency, against the same agreement, and it matches without help. The exceptions are generated by structure, not by sloppiness. The largest public consolidation in the United Kingdom names three reasons a counterparty pair fails to agree: a legitimate difference of opinion on how to treat a transaction, incomplete information in one counterparty, and an error made in the collection tool. Only the third is what most people mean by a reconciliation error, and it is the only one of the three a matcher can close by itself.	HM Treasury, Whole of Government Accounts 2023-24, paragraph 3.21
Volume	Not transaction count. Effort scales with the number of entity pairs, currencies and charts of accounts, because each pair is its own agreement with its own cut-off and its own dispute. The scale is real: 21,605 United States corporation returns reported 106,732 controlled foreign corporations for tax year 2021, and receipts of those foreign corporations from related parties were 2.59 trillion dollars against 9.26 trillion dollars of receipts from all sources, with payments to related parties 3.91 trillion dollars of 9.09 trillion dollars. The payback condition follows: where a group is one country, one currency and one chart of accounts, and every intercompany pair fits on one page, the matching is not the cost and nothing here pays back.	IRS Statistics of Income, Controlled Foreign Corporations, Tax Year 2021
Cost of an error	Two-tier, and the first tier lands somewhere other than group profit. A pair that does not agree does not vanish; the residual sits in the consolidated statements as an unexplained difference and an auditor asks about it. The public instance is instructive: a consolidation of over 10,000 public bodies was qualified on lack of evidence supporting intragovernment eliminations until the 31 March 2016 balance sheet, and the preparer names preventing that qualification from returning as one of the key risks it manages. In the 2023-24 accounts the residual elimination uncertainty was 1.7 billion pounds of debits and 2.1 billion pounds of credits on the balance sheet, and 2.9 billion pounds of debits and 2.5 billion pounds of credits on the statement of revenue and expenditure. The second tier is tax, and it is assessed years later.	HM Treasury, Whole of Government Accounts 2023-24, paragraphs 1.57 and 3.22
Reversibility	High inside the period, one-directional across the consolidation lock, and on the tax side not a reversal at all. When a tax authority adjusts one member's income, a correlative allocation to the other member follows, and it must be reflected in that member's documentation whether or not it changes any liability for an open year. Conforming the cash accounts afterwards can mean treating the adjustment as an account receivable due as of the last day of the year of the transaction. So an intercompany difference resolved three years later does not undo an entry; it adds two more, in a period everybody closed.	26 CFR 1.482-1, paragraphs (g)(2) and (g)(3)
Regulatory exposure	Named in the standards by process, and the same number is filed twice on two bases. IFRS 10 requires elimination in full and caps the gap between a subsidiary's reporting date and the group's at three months; IAS 21 states that an intragroup monetary item cannot be eliminated against its counterpart without showing the results of currency fluctuations. Separately, a group above 750 million euro of consolidated revenue files a country-by-country report that splits revenues into related party and unrelated party for every tax jurisdiction, within twelve months of the year end, and the instructions say plainly that it is not necessary to reconcile that report to the consolidated financial statements. Two filings, one underlying transaction, and no obligation to make them tie.	IFRS 10 B86 and B93; IAS 21 paragraph 45; Council Directive (EU) 2016/881, Annex III
Vendor market maturity	Mature, and bundled into the accounting package or the consolidation tool rather than bought beside it. No public measure of market concentration in this category was retrieved, so none is quoted here. What is documented is the shape of the shipped capability: one incumbent's product documentation describes elimination rules keyed on a trading partner financial dimension, net change or fixed elimination methods, an elimination proposal a person reviews before posting, consolidation across different charts of accounts and different fiscal calendars, and currency translation into a dedicated consolidation entity. The same documentation records two limits worth knowing before buying: minority interest is manual calculation and entry, and the elimination journal shows amounts in the currency of the originating transactions rather than the accounting currency, which it warns users may find confusing.	Microsoft, Dynamics 365 Finance consolidation and elimination documentation

Two of those rows carry the argument. The exception-variance row says the residual population is manufactured by the group’s own structure, which means it is knowable in advance rather than discovered each month: a group with four currencies, two charts of accounts and one subsidiary on a September year end knows exactly which pairs will disagree before any software runs. The regulatory row says the two filings are not required to agree with each other, so a system built to make every intercompany number identical everywhere is solving a problem the rules do not pose.

The reversibility row is the quiet one and it sets the design. Inside the period a difference costs an entry. After the consolidation is locked it travels forward, and on the tax side it can arrive years later as a correlative allocation that has to be documented on both sides regardless of whether anything is payable. That is an argument for a matcher that runs continuously and exposes an open pair the day it appears, and against anything that batches the comparison into the last week of the close.

The reconciliation is a policy question wearing a data problem’s clothes

The most quoted number about intercompany work is how long it takes, and the largest public example is worth reading carefully because it separates the two things that make it long. The United Kingdom consolidates over 10,000 designated public bodies; for 2023-24 that meant 657 collection returns requested and 456 received. The preparer’s own comparison of its accounts with the National Accounts describes the difference in one line: transactions within government “are eliminated entity-by-entity in a process that takes several months”, where the faster method’s residual elimination error cannot even be quantified. Several months is the honest figure, and it is for the elimination alone.

Automation moved that. Publication of the accounts went from 20 months after year end for 2009-10 to 15 months for 2012-13 and 12 months for 2013-14, and the preparer attributes the improvement to improved eliminations processes. The audit qualification on intragovernment eliminations, which had run from the first published accounts, came off at the 31 March 2016 balance sheet as the systems matured. Those are real returns and they are the returns to expect: a shorter calendar and an opinion that stops mentioning the eliminations.

What did not move is instructive. The accounts for 2023-24 were still disclaimed by the Comptroller and Auditor General, and the reason was not the elimination arithmetic. It was missing and unaudited data from bodies that did not submit. A counterparty that does not report is not an unmatched pair; it is an absent one, and the consolidation discloses it as a related party transaction instead. No matching engine reaches it.

The same shape governs a company group at a smaller scale. The pairs that never agree are the ones where the two entities never agreed a rule: which side raises the invoice, which cut-off applies, which currency the balance is carried in, who arbitrates a disputed recharge and by when. A group that writes that down first automates the agreement. A group that does not automates the disagreement, and gets its disputes sooner and in a better format.

What automating intercompany reconciliation is not

It is not the elimination itself. Once a pair agrees, removing it is deterministic, and it has been shipped in the accounting package for years: rules against a trading partner dimension, a net change or fixed method, a proposal somebody reviews, posting into a designated elimination entity. Paying to rebuild that is the most common thing found underneath an intercompany automation request, and the configuration is a week rather than a project.

It is not a currency fix. The standard is explicit that an intragroup monetary item cannot be eliminated against its counterpart without showing the results of currency fluctuations, because the item is a commitment to convert one currency into another and it genuinely exposes the group to a gain or loss. A system that nets those differences to zero is not reconciling; it is misstating. What automation can honestly do here is translate consistently and show the exchange component of a difference separately from the transactional component, so that the person reviewing knows which part is required to be there.

It is not a tax position. The country-by-country template may be built from consolidation packages, from separate statutory accounts, from regulatory statements or from internal management accounts, provided the source is used consistently, and no reconciliation to the consolidated statements is required. Groups therefore file related party revenue on a basis that need not equal the intercompany revenue eliminated in the group accounts, and that is compliant rather than careless. Building a system on the assumption that one intercompany number exists creates a discrepancy where the rules created a distinction.

And it is not a substitute for a group policy. The one input that changes the size of the residual population is the set of rules the entities have agreed with each other. Everything else changes how quickly the residual appears.

The verdict

The evidence supports automating the matching early and running it continuously, and supports expecting the return in the calendar and in the audit position rather than in headcount. The profile is what settles it. The exception-variance row says the exceptions are structural, so they can be enumerated before a line of anything is built, and a group that cannot name its own reconciling categories is not ready to buy. The volume row says the cost scales with pairs, currencies and charts of accounts rather than with transactions, so the payback condition is structural too: several entities, more than one currency or chart of accounts, and a fixed group reporting date. Where a group is one country, one currency and one chart of accounts, and the accounting package already carries elimination rules, there is no automation problem and the correct first purchase is the configuration already licensed.

The reversibility and regulatory rows decide the shape of what gets built. Because a difference costs an entry inside the period and a documented two-sided adjustment years afterwards, the value sits in continuous exposure of open pairs rather than in a faster comparison at the end. Because the standards require some differences to persist and the tax filing is explicitly not required to tie to the group accounts, the system has to be able to say which category a difference belongs to, not merely that one exists. A reconciliation tool that reports a number without saying whether it is a cut-off, a currency effect, a policy divergence, a tax basis or a genuine error has automated the arithmetic and left the entire job.

The honest expectation to hold, then, is a shorter and better-evidenced consolidation with the same number of arguments in it, until the group writes down who owns each intercompany relationship and what happens when the two sides disagree. Foxnut Studios builds systems of this kind and hands them over, which is a reason to be direct about the case where the answer is no: a group whose intercompany balances differ because two finance teams have never agreed a cut-off does not have a matching problem, and buying a matcher will tell it that in higher definition and at greater expense than a meeting would.

If you are not sure which of those two your own group has, describe the disagreement to us before buying either a matcher or a meeting facilitator.

Sources

Foxnut Studios works on briefs like this one from Bengaluru and Paris. If you want the shape of that before you talk to anyone, here is how an AI engagement is scoped and priced.