A television on a company invoice can prompt an obvious question: is this a business purchase or private consumption? Electronic invoicing makes the line easier to find. It does not put the television in a particular room, establish who uses it or explain the business purpose.

That distinction is not a criticism of structured invoices. Better records can help an accountant reconcile transactions and help an authority select questions. The problem begins when a clean digital record is treated as a complete account of economic reality.

We should neither dismiss the invoice as worthless nor promote it into evidence it cannot contain. Good tax administration needs both structured information and a disciplined understanding of its limits.

Several different claims hide inside one document

An invoice normally records what a supplier says it supplied, to whom, when and for what amount. A structured message makes those assertions easier for software to process. A delivery acknowledgement can provide evidence about a message reaching a system. Neither event logically proves that the stated goods were delivered, that a service was performed or that the purchase has the claimed tax treatment.

These are different evidentiary layers. First comes document integrity: is this the record that was sent? Then comes attribution: which account or organisation submitted it? Next comes the underlying transaction: did the supply occur? Finally comes the tax conclusion: what follows from the transaction under the applicable rules?

A successful technical validation belongs primarily to the first layers. Even a perfectly formatted invented transaction would still be invented. Conversely, a genuine transaction can have a defective description. The system should not confuse either situation with the other.

A television does not have one inevitable business meaning

Consider a hypothetical company buying a television and a washing machine. It might operate guest accommodation, install equipment for customers or maintain demonstration facilities. The same product categories could also appear in an improper private purchase charged to a business. These are possible scenarios, not observations about any particular taxpayer.

The invoice line alone does not distinguish them. Useful follow-up evidence might include the purchase order, delivery destination, asset register, installation record and explanation of actual use. Which records are necessary depends on the question; the point is not to demand every possible document for every purchase.

Suppose the business says the television went into a reception area. The claim can be checked against other evidence. The product name can legitimately trigger a question, but it cannot answer the question by itself. A rule that automatically treats all televisions as private consumption would replace factual inquiry with a category assumption.

A construction company makes the example sharper. Imagine it buying twenty televisions and five washing machines: screens might support site monitoring or presentations, while machines might wash work clothes. These possibilities prove neither actual use nor VAT entitlement; they must be supported by the facts. They rebut an automatic conclusion from the product name and quantity, not the need for evidence. The quantities can prompt reasonable questions about locations and use; they do not settle those questions. Nor does a business name on the invoice establish that the claimed use occurred.

Services expose the limitation more sharply

Goods at least have identifiable physical characteristics. Services may be described by a period, a task, a contractual milestone or an agreed scope. “Services under contract” can be too vague to tell a reader what was supplied, while a full narrative may reveal confidential material unrelated to routine tax processing.

There is therefore a genuine design problem between inadequate description and excessive disclosure. An invoice should convey the information legally required to identify the supply. Supporting documents can provide additional context when properly requested. A national data collection should not assume that copying every contract and deliverable into each invoice is the only way to make services auditable.

For a hypothetical software maintenance contract, a useful description might identify the maintenance service and relevant period, linked to a contract reference. The underlying agreement, service records and acceptance correspondence would then explain the work. This is an illustration of evidence design, not a guarantee that one wording satisfies every national invoicing rule.

An example of an overly brief description—not a template to use—is “supply of goods under contract no. XXX, amount Y.” It illustrates the gap between an invoice line and the transaction behind it. It is not a recommended universal invoice description: a contract reference alone may fail to identify the quantity and nature of goods or the extent and nature of services sufficiently.15 Nor would it defeat all analytics; parties, amounts, timing and relationships remain available signals. The question is what can fairly be inferred from them.

A personal banking example—and its limits

In my own experience in Spain, banks asked me to document larger incoming or outgoing payments with an invoice, a contract and sometimes delivery or transport documents before the payment could proceed. I describe this as a personal experience of anti-money-laundering checks, usually abbreviated AML, not an independently audited account of every Spanish bank’s practice.

The relevant point is that the invoice was not always the end of the inquiry. Supporting documents helped explain the underlying transaction. This does not mean banks replace tax authorities, that AML and tax tests are identical, or that a bank’s acceptance proves VAT deductibility. It shows why a document that initiates a check need not be sufficient to conclude it.

Three questions must stay separate: how the purchase is recognised in the accounts; whether and when its cost is deductible for income/profit tax, including any depreciation treatment; and whether input VAT is deductible. The following judgment addresses the third question, in the specific context of invoice descriptions for services. It does not decide the accounting or income-tax treatment of the televisions.

The Court of Justice addressed this distinction

In Barlis 06, case C-516/14, decided on 15 September 2016, the Court of Justice considered invoices for legal services and the refusal of input VAT deduction because their descriptions were insufficient.15 The case is valuable precisely because it does not offer a simplistic answer in either direction.

The Court explained that Article 226(6) of the VAT Directive requires the extent and nature of services to be specified, but does not require an exhaustive description of the particular services supplied.15 It considered that the general descriptions at issue did not, on their face, satisfy the relevant requirements, subject to the national tribunal’s assessment.15

At the same time, the Court held that the authorities could not refuse deduction solely because the invoice failed those formal requirements when they had all the information necessary to determine whether the substantive conditions were satisfied.15 Paragraph 44 expressly says that authorities cannot restrict themselves to examining the invoice: they must also consider additional information provided by the taxable person.15

This is not permission to ignore invoice requirements. The judgment also recognises the possibility of penalties for failures to meet formal conditions.15 Nor does it grant deduction for a fictitious supply. Its practical lesson is that a defective description and the absence of a real entitlement are different questions.

More data do not automatically close the evidence gap

One response is to demand richer invoice lines. That can help where the problem is genuinely missing identification: an intelligible service category and a defined period are more useful than an empty label. But additional fields cannot, by themselves, establish that an assertion is true.

Another response is to collect contracts and supporting files centrally. That may make some inquiries easier, but it creates a separate decision about necessity, access and retention. Where personal data are involved, the requires purpose limitation and data minimisation; these are not displaced merely because a dataset might later be useful.21

Commercial confidentiality also deserves design attention, without pretending that every confidential company fact is personal data. The GDPR protects natural persons; its recital 14 distinguishes information concerning legal persons.21 A price list or technical method may need protection for reasons other than data-protection law.

My proposal is to keep routine invoice exchange focused on its purpose and make additional evidence available through an authorised, proportionate process. That requires good recordkeeping by businesses and competent requests by authorities. It does not mean that supporting information should be inaccessible to legitimate control.

The strongest counterargument: scale requires signals

A tax administration cannot investigate every transaction from the beginning. Product categories, unusual relationships and inconsistent amounts can be useful ways to prioritise scarce attention. Requiring complete proof before even asking a question would make risk-based control impractical.

I agree with that objection. The boundary should be between selection and conclusion, not between using data and refusing to use them. An invoice can justify a targeted request without establishing the eventual finding. The request should explain what needs clarification rather than present an untested suspicion as settled fact.

There is also a cost to the business. If every unusual but lawful purchase generates the same extensive questionnaire, the control design may create substantial work without improving discrimination between cases. Authorities should test whether their selection rules identify relevant issues and whether the follow-up burden is proportionate. No empirical success rate is assumed here; measuring it is part of the proposal.

A better evidence workflow

For businesses, the practical response is not to write a defensive essay on every invoice. It is to use meaningful descriptions, preserve links to contracts and orders, record receipt or performance where appropriate, and retain corrections without destroying the relationship to the original document.

For system designers, message status should be labelled accurately. “Validated,” “delivered” and “accepted by the customer” should not silently become interchangeable. Acknowledgements should identify the event they record and preserve enough context to reconstruct it. For example, a provider’s exportable record of submission, rejection and corrected resubmission could reconstruct a delivery dispute. That is valuable evidence of technical events, not proof that the contracted service occurred.

For tax authorities, a risk signal should lead to a specific evidence question. A decision should distinguish the invoice, supporting material and reasoning. Investigate use when use is disputed, and performance when performance is disputed. Neither a public channel nor a commercial provider’s delivery record closes that evidentiary gap.

Electronic invoices can make evidence easier to organise, transmit and reconcile. They cannot abolish the difference between a statement and the reality it describes. A credible reform will be stronger, not weaker, when it admits that boundary and designs its controls around it.

Sources

[15] Court of Justice — Barlis 06, C-516/14, judgment of 15 September 2016

[21] Regulation (EU) 2016/679 — GDPR, especially Articles 5, 22, 25, 26, 28 and 32