There is a moment in every continuous reporting regime when a finance team realises something uncomfortable. The tax administration now holds a version of the company's VAT position that the company did not prepare, did not review, and cannot see in full. It was assembled from documents as they were issued, by a process nobody in the business supervises, and it exists whether or not anyone has looked at it.
The return, meanwhile, is prepared the way returns have always been prepared: from the ledger, at period end, by people applying judgement to figures that have settled.
Two versions of the same quarter, built from different sources at different moments by different methods. They will differ. The only question is whether the business knows by how much before somebody else asks.
Three records, not two
It helps to be precise about what is being compared, because "the data" is three distinct things.
There are the documents you issued: individual invoices and credit notes, each with a date, a counterparty and a treatment. There is the data that was reported: whatever left your systems and reached the administration, which is derived from those documents but is not identical to them, because reporting has its own scope rules, its own timing and its own transformations. And there is the return: a set of aggregate figures produced from the ledger, which includes things that were never invoiced at all and excludes things that were.
Reconciliation is not one comparison. It is two, and they fail differently. Documents against reported data is a completeness and transmission question. Reported data against the return is an accounting question. Teams that run only the second are the ones surprised by a document that never arrived.
The four kinds of break
Every difference worth investigating falls into one of four categories, and knowing which one you are looking at determines who fixes it.
| Category | What happened | How it presents | What closes it |
|---|---|---|---|
| Timing | The document and the ledger entry fall either side of a period boundary | A break in one period, offset by an equal and opposite break in the next | A schedule showing the reversal; no correction is needed |
| Scope | Something is in one record and legitimately not in the other | A persistent difference of a similar size every period | A documented list of what is deliberately excluded and why |
| Correction | A credit note, cancellation or resubmission landed in a different period from the document it corrects | A break that moves rather than clears | Linking the correction to its original and showing the net position |
| Genuine error | A document was never reported, reported twice, or reported with wrong figures | An unexplained residual after the other three are removed | A correction to the reported data, and a control preventing recurrence |
The order in that table is the order to work in. Strip out timing first, then scope, then corrections. What remains is the only part that needs anyone senior to look at it, and it is normally small. Teams that attack the residual first spend weeks investigating differences that were always going to reverse.
Why the residual is the interesting number
An administration running continuous reporting is not primarily looking for arithmetic errors. It is looking for patterns: a supplier whose reported outputs consistently exceed declared outputs, a counterparty whose purchases do not correspond to anybody's sales, a period where the shape of the data changes.
The residual after timing, scope and corrections is what that analysis sees. If it is small, stable and explainable, the business is in a defensible position even when individual documents turn out to be wrong. If it is large or volatile, the individual documents barely matter — the pattern is the finding.
This is a genuine change in what compliance means. The old test was whether the return was right. The new test is whether the return is consistent with everything the administration already holds, and consistency can fail while correctness holds.
The regimes make it worse before they make it better
Different national arrangements collect different things, which means a business operating in several Member States is not running one reconciliation but several, against records that do not agree with each other either.
A ledger reporting regime collects book entries on a short clock, so its record resembles your ledger but arrives before the ledger has settled. A real-time invoice reporting regime collects the document at issue, so its record resembles your billing system and knows nothing about later accounting adjustments. A standard audit file is generated from the ledger on demand, long after both, and is the artefact most likely to contradict the other two because it reflects a position that has since been corrected.
Three records of the same transactions, from three systems, at three moments. All three are supposed to be true. Reconciling to any one of them individually is straightforward. Reconciling to all three, and explaining why they differ from each other, is the real work, and it is not made easier by the fact that the differences are usually legitimate.
The commonest arrangement is no arrangement: nobody reconciles until an administration asks a question, at which point somebody reconstructs eighteen months of differences under time pressure, using systems that have since been upgraded. The reconstruction is always worse than the contemporaneous record would have been, and its weakness is not the figures. It is that it was built to answer a question rather than to be true.
Making it a control rather than an exercise
The difference between a reconciliation and a control is that a control has an owner, a frequency, a tolerance and a documented outcome — including the outcome "no exceptions", which is the one people forget to record.
That is not bureaucracy for its own sake. Under the business control route to assuring an invoice, the reconciliation is part of the audit trail linking the invoice to the supply, and an undocumented control is indistinguishable from no control when somebody asks you to demonstrate it. A reconciliation that is performed diligently every month and never written down protects the business operationally and not at all evidentially.
The evidence requirement is modest: what was compared, over what period, what differences were found, how each was categorised, and what was done. A page a period. What makes it valuable is that it was written at the time.
What the reforms change, and what they do not
The convergence programme in the VAT in the Digital Age package will reduce the number of national dialects a multinational has to reconcile against. It will not remove the reconciliation itself, because the underlying condition is not divergence between regimes — it is that transaction data and accounting data are different things assembled at different times, and both will continue to exist.
Anyone hoping that harmonised reporting will make this go away has misread the problem. It makes the comparison more uniform. The differences remain, and the obligation to explain them remains with the taxable person, exactly where it has always been.
