Budget Audits: Finding Missing Units, Quantities, and Prices Before They Hurt Margin

Most estimate errors are not dramatic. They are quiet omissions: a unit left blank, a quantity copied from the wrong room, a price still set to zero, or a chapter that looks complete but contains no priced work.

None of these throw an error. The spreadsheet still sums. The PDF still prints. The total at the bottom looks plausible because a missing line contributes nothing, and nothing is invisible. The damage shows up later, on site, when the work is real and the number that was supposed to cover it was never there. That is why these omissions are so expensive: they pass every check a budget normally gets because the only check that catches them is a deliberate, line-by-line audit, and that is exactly the work nobody has time for under bid pressure.

Omnicost treats those gaps as first-class signals. Budget rows are expected to carry a name, unit, quantity, unit price, and total. When a line item is missing one of those fields, the agent can identify it, search the catalog, apply a reasonable unit, or flag the item for review.

The order matters. The agent does not guess silently. If a row has a name and a quantity but no unit, it looks at the description and the magnitude of the quantity to propose a unit — square metres for a finish, linear metres for a skirting, units for fixtures — and shows that proposal as a suggestion, not a fact. If a row has a unit and quantity but a price of zero, it can pull a candidate from the catalog and mark where the number came from. If it cannot resolve the field with reasonable confidence, it stops and flags. A flag is not a failure; it is the system refusing to invent a number you would later have to defend.

This is especially useful after an estimate has been generated quickly from a brief. The first draft gives the team structure. The audit pass turns that structure into something usable. It separates priced rows from placeholders, checks totals, and removes empty wrappers that make a budget look more complete than it is.

A pass over a real budget

Say you import a budget and run an audit over it. Three things surface.

First, a tiling line: "Porcelain floor tile, bathrooms," quantity 48, but the unit cell is blank and the price reads 0 €/m². On a standard review this row reads as harmless — it adds zero to the chapter total. The agent does the opposite of harmless: it sees a finish description with a quantity of 48, proposes m² as the unit, searches the catalog for porcelain floor tile, and surfaces a candidate unit price with its source. You confirm or override. Either way the line now carries real money instead of a quiet zero.

Second, a structural chapter that looks complete — eight rows, headings, sub-headings, the lot — but when the agent separates priced rows from placeholders, every row underneath is a description with no quantity and no price. It is a heading dressed as work. The agent flags the whole chapter: this is scope that has been named but never costed. That is the difference between a budget that looks finished and one that is.

Third, an empty wrapper: a parent grouping whose children were all deleted or never added. It contributes nothing and inflates the apparent structure of the document. The agent removes it, so the budget you send reflects the work you actually priced.

The goal is not to replace estimator judgment. It is to make review faster and more systematic. A human should decide whether the assumptions fit the project. The system should make sure the assumptions are visible.

This matters most when budgets move between formats. A BC3/FIEBDC-3 export, a Presto file, or a spreadsheet each lose or mangle fields differently in transit — a unit code that does not map, a price decomposition that flattens, a chapter record that arrives without its children. The audit pass is where those translation losses become visible items instead of silent gaps carried into the bid.

Over time, audits become stronger as more catalog data, supplier observations, and project history enter the system. Omnicost can compare a line item against recent market prices, regional medians, and similar past budgets.

That comparison turns the audit from a completeness check into a sanity check. A field can be present and still wrong: a price that is technically filled in but sits far below what the same work cost on three recent budgets is worth a second look before it ships. The system surfaces the gap; the estimator decides whether it is a genuine deal, a stale number, or a typo.

Margin is often lost before work begins. A budget audit gives teams a practical checkpoint before a quote leaves the building.