Five Rates, Two Outcomes: Irish VAT and the Invoice Line

Ireland has five rates of VAT running at the same time: a standard rate of 23%, a reduced rate, a second reduced rate, a livestock rate of 4.8%, and a zero rate. Which one applies to a sale is a determination about what was sold — not about who sold it. That is an unremarkable fact about Irish trade and an awkward one about most software, because most software treats a tax rate as a setting on the company. This is about the point where those two ideas meet, which is a single line on an invoice.

Nothing here is tax advice. Which rung a particular supply sits on is your accountant's determination and we would not offer a view on it — and we have deliberately not printed a list of what belongs on which rung, because the rungs are stable and the assignments have moved more than once. What follows is about data modelling, which is a question you can settle without an adviser.

The question that decides everything, and it takes one sentence

Ask a vendor whether they support multiple VAT rates and you will get a yes, because it is true in some sense of every system ever written. The useful question is narrower: which record is the rate a property of? There are only three plausible answers and they are nowhere near equivalent.

• The organization. One default rate, set once, applied to everything unless overridden. Entirely correct for a business that sells one kind of thing.
• The item. Each product or service carries its own rate, so a mixed invoice comes out right because the items know what they are.
• The line. The rate is settled when the sale is recorded — the only answer that survives a supply whose treatment depends on how it was bundled with others.

Ireland needs the second answer as a floor and the third for the awkward cases. A great many systems give you the first and describe it as the second, because an item can be flagged exempt and that feels like item-level tax. It is not the same thing, and the space between them is where Irish invoices go wrong.

What our own product does, stated exactly

We will use ourselves as the worked example, because it is the schema we can actually show you and because a post on this subject that only criticised other vendors would be worth nothing. Our resolver takes an organization default rate and an item treatment. It has three branches.

Three branches, two outcomes. For an Irish organization whose default is the standard rate, the top rung of the ladder and the bottom rung are reachable and the three in between are not — except by making one of them the whole organization's default, which is right for a single-rate business and wrong the moment one line differs from the rest.

Why a wrong rate is harder to find than a missing one

This is what makes rate modelling different from most data problems, and it is the reason the failure survives so long in practice. A missing figure announces itself: a report has a hole, a total does not tie, somebody asks. A figure computed at the wrong rate does none of that.

It balances. It reconciles. Subtotal, tax and total are internally consistent, because all three were computed from the same wrong percentage. Every arithmetic check you own will pass, and so will your accounting system's, because it is being handed a coherent document.

What finds a wrong rate is a person who knows what the rate should have been, reading documents one at a time. That is not a control. That is luck, and it scales in exactly the wrong direction.

And the exposure grows with the thing you would most want to grow. The businesses issuing the most mixed invoices — the ones selling a service with goods in it, or goods with installation attached — are the ones for whom a single-rate model is most wrong, and they are usually the ones with the most invoices.

Two more things Ireland asks that a rate column cannot answer

Zero-rated and exempt are not the same thing

Both produce nil tax on a line, which is why systems collapse them into one state so often. Ours does. They are genuinely different: a zero-rated supply is taxable at a rate of nought, and an exempt supply is outside the charge. The consequences differ for what a business can recover, so the distinction has to survive being saved if a return is ever going to be produced from the records rather than reconstructed alongside them. In our schema it does not survive — both resolve to zero, and afterwards the record cannot tell you which was meant.

A mixed supply can change the rate of the whole thing

Ireland has a two-thirds rule: where a contract is a service but the goods within it exceed two thirds of the total value, the supply can be taxed as goods rather than as a service. Read that as a mechanism rather than as a formula to implement — whether it bites on a particular contract is your adviser's call. What matters architecturally is the shape of the question, which is a test on the composition of a supply. The answer depends on the other lines. No setting above the line can produce it, and neither can an item-level rate, because the item does not know what it was sold alongside.

Four questions worth asking any vendor

What we would build, and where we stop on purpose

The build is a rate set the organization defines, a selector on the line, and a stored distinction between zero-rated and exempt. It is a schema and resolver change rather than a configuration exercise, and it would improve every multi-rate market in the product rather than only Ireland — which is the argument for doing it properly rather than for one country. What we will not do is tell you which rung a supply belongs on. That is a determination your accountant makes and signs, and a software vendor with an opinion on it is offering reliance it cannot carry.