Twelve per cent

According to the Central Statistical Office's release of 14 September 2026, construction output in Hungary was 12.2 per cent lower in July 2026 than a year earlier. Buildings were down 10.7 per cent, civil engineering works 14.0 per cent. The trade had already felt this; now there is a number for it.

Two things belong with that figure before anyone turns it into a headline. The 12.2 covers a single month: across January to July the fall is 2.7 per cent. And the order book at the end of July was, in aggregate, 0.8 per cent larger than a year before — reassuring on first reading.

Less so on the second. The same release breaks it down: the order book for building construction shrank by 1.9 per cent, while the one for civil engineering grew by the same amount. The aggregate gain is carried by roads, rail and utilities. If you build buildings, your backlog did not grow. It shrank.

The real number sits with new orders: the volume of new contracts signed in July was 24.8 per cent below last year, and for building construction 29.5 per cent. The backlog is not what disappeared. The pipeline is.

01 — MarginWhat last year still absorbed

Fewer new jobs, the same number of firms chasing them, and the effect shows up in exactly one place: the bid price, and then the margin built into the contract sum. These items are the same size this year as last. Last year they still fitted inside the margin:

• the work that was always in the technical scope but missing from the priced bill of quantities — under a lump-sum contract you are not paid for it, even though you paid for it;
• the variation you would rightfully be paid for, but which started on a verbal instruction: with no entry in the statutory construction log and no signed order, there is nothing to certify it with;
• the time-related site costs that ran on through the two-week delay — site setup, supervision, site cabin, crane hire;
• the stripping out and rebuilding that followed from you and your subcontractor both reading the technical scope in good faith, and differently;
• the retention you should have reclaimed when the warranty period ran out — except the project had long closed, and nobody was keeping the date.

This year the same list comes straight out of the result.

In a tight market the winner is not the one who bids lowest. It is the one who knows what they bid.

02 — The bidWhat the submission deadline decides

Bidding turns into betting in a market like this. The deadline is short, the tender documents are incomplete, the technical scope is open to reading in several places — and the unit rates still have to be entered by Thursday noon.

What you would need in order to decide is roughly this: what the actual unit rate came to for the same trade on the last two projects; how far the subcontractor you are about to price slipped; how much extra work they submitted, and how much of it you had to accept.

That data exists inside the company. In the contracts of a closed project, in the annexes to signed performance certificates, in an email thread from the spring. It is not missing — it simply does not assemble itself into a single table before the deadline passes.

The last two days are not consumed by your own spreadsheets in any case, but by incoming subcontractor quotes and the tenderer's supplementary information. Which is precisely why those two days should go on what they are for: comparing the quotes and setting the risk allowance. Not on someone digging out what the same trade cost last year.

03 — The loadWhat converges on the owner

The administrative load rises meanwhile. The same amount of work takes more bids. There are more omitted works and more variations, because the client is economising too and reshapes the technical scope as the job runs — the contract sum falls, while your site setup, your supervision and your crane stay exactly where they were.

Payment slows down too, though not where you are looking for it. The clock starts at the performance certificate, so anyone who wants to defer does not sit on the invoice — they sit on the signature: a comment, a discussion, one more round. Your overdue receivables are nil on paper, and not on the bank statement.

Each of these steps is justified on its own; none of them is waste. The problem is their sum, and the fact that they all converge on the owner — not from a wish to control, but because there is nowhere else for them to go. You can delegate what has somewhere to land.

04 — Five sourcesData that lives in five places

Consider where a project's data sits today at an average contractor.

• The contract: in Word, in a shared folder, in two or three versions, with the amendments in a separate file.
• The performance certificate: a PDF attached to an email. The subcontractor's sits with your project manager, yours with the client's site engineer — and often enough one side is still working from the previous version of the measurement.
• Material, plant-hire and skip invoices: with the bookkeeper, with no project or trade marked on them. The specialist subcontractor's invoice you can still identify; the twenty kinds of material invoice only by hand.
• Margin: in a spreadsheet one person maintains, and in which nobody else dares touch a formula.
• Defects and variations: in a chat group, with photos, undated — and the variation itself in two people's differing recollections, with no signed order.

When a management report is needed out of this, somebody sits down and merges it. Over several days, after chasing three colleagues. A number comes out at the end, and if anyone asks where it came from, the answer is another spreadsheet. A hand-assembled report has no line you can point at and ask which invoice exactly did this come from.

05 — The processWhere the data is entered once

ATMOS was built so that those four questions stop being research projects. The process stays in one piece: tender, contract, performance certificate, invoice, payment — each step working from the one before.

The winning bid from a subcontractor tender becomes a draft contract in one step: with the quoted total, the payment terms, the advance, the performance and warranty retention, and the dates. The client-side performance certificate is assembled from the line items of the project budget: you enter a monthly percentage or amount per line, the system keeps the cumulative progress and calculates retention — performance retention on every interim certificate, warranty retention at final account, where the earlier retention is released. The outgoing invoice is issued from the signed certificate.

Incoming invoices arrive from the tax authority's data service, with no typing. The partner is identified by tax number; if exactly one performance certificate matches — same partner, the amount within a hundred forints, the delivery date inside the certificate's period — the project and the contract are attached automatically. The trade and the budget line you set yourself; that is not a prediction, it is a dropdown with that partner's lines at the top.

Three things make this work, and all three are duller than they sound. An approval order: who approves what, in what sequence, and what happens when they do not. Structured data: project, trade, partner, date, amount — in fields, not in the middle of an email. And a trail: signatures, deductions and variation orders with dates on them rather than recollections.

Two things should be said plainly, because the first day of a trial would reveal them anyway. ATMOS does not replace the statutory electronic construction log — there is no data connection to it; the final-account checklist simply requires you to confirm by hand that the log has been closed. And it does not draw the statutory line between unpriced scope and ordered variations for you: there is one process for extra work, with a direction and a reason code. Which of the two a given item is remains a matter of the contract and the bill of quantities.

06 — SpeedThe report that is already there

The biggest difference is what stops happening. A management summary no longer has to be commissioned, chased and then believed. It has to be opened. Whoever used to merge the spreadsheets has not had their job taken away — they simply no longer produce a number by hand from five sources that nobody can verify afterwards.

And in a tight market a fast decision is money:

• the deduction you apply before the certificate is issued, rather than afterwards, in an argument;
• the subcontractor who never makes it onto the invitation list, on the strength of their own earlier ratings;
• the margin slipping, visible when the subcontract is signed — while there is still a decision to make — not at the third invoice;
• the interim certificate you can submit on the turn of the month, because the line items are already together — not two weeks later, with the payment clock sliding by the same amount.

None of these is a big decision. In a thin year they nonetheless decide whether anything is left at the end of the project.

07 — The costFive pieces of software, or one

The classic answer to this job is five separate tools: drawing register and defects, contracts and performance certificates, invoicing with a tax-authority link, document storage and corporate email, and financial controlling. Each is good at its own task. The data, however, does not move between them on its own — you are the transfer mechanism.

What those five subscriptions come to can be worked out: the vendors' list prices are public. The calculator on the site holds the September 2026 list prices with a clickable source on every line; you supply the headcount, the rest is multiplication.

What the calculation leaves out is the more expensive half: the work that arises between the five systems. Work out once how many hours go, at your company, into getting the monthly report to stand up — chasing and reconciling included. In most places that is not hours but days. That line appears on no price list. You pay it every month all the same.

One question of size should be settled, since it would come out anyway. There is a company size below which separate tools are cheaper — and headcount is not what decides it. What decides it is how many projects run at once and how many live subcontracts you hold. By our own calculation the tipping point sits at two or three concurrent projects and around eight office staff. A four-person firm running one project should not be buying this.