The hidden cost of lost consultations: the expense that no one calculates.

A real cost item, treated as an overhead expense

Reading the technical specifications, taking measurements, consulting suppliers, preparing the detailed price schedule, technical report, assembling the file: responding to a consultation requires engineering hours that the company pays for in full, whether it wins the contract or not.

In most companies, this time is nowhere recorded. It isn't allocated to a specific project because there are no projects. It ends up diluted in overhead costs, just like rent. As a result, the manager knows precisely the profit margin on their projects but has no idea of ​​the unit cost of producing a file.

This is all the more striking given that it's the only area where the company invests qualified technical resources without any guarantee of return. The success rate, however, is tracked everywhere. But a success rate says nothing about the cost. Two companies with a 25% success rate can have radically different savings depending on the cost of each response they produce.

What the economic outlook says for 2026

The current context does not support leaving this position in the shadows.

The French Building Federation anticipates a 1,8% increase in activity in volume for 2026, driven by new housing (+9,5%), after a 4% decline in 2025 — a rebound that does not make up for the previous year, and a virtually flat non-residential sector (+0,5%).

On the public works side, the National Federation of Public Works forecasts a contraction of 3,2% in volume, due to the combined effect of the municipal cycle and budgetary decisions of local authorities.

Failures remain high: according to Altares, the construction sector recorded 4.148 collective procedures in the second quarter of 2026, a figure that is almost stable (-0,8%), but which masks very contrasting situations between structural work, public works and real estate development.

At the same time, the volume of public procurement remains considerable. The Public Procurement Economic Observatory recorded 223.383 public contracts in 2024, totaling €233,3 billion, or nearly 8% of GDP. SMEs win 60% of the contracts by number, but only 25% by value.

One point deserves mention: no French public statistics track the average number of candidates per tender. The OECP's data relates to awarded contracts, not submitted bids. In other words, the entire sector's response effort—thousands of hours of research per week—is not measured by anyone. Each company is therefore left to rely on its own figures.

The calculation framework

It fits in four lines and uses only data already present in the company.

Real-time time per project, step by step: qualification of the tender documents, review of the technical specifications, quantity surveying, supplier consultation, preparation of the bill of quantities, technical proposal. Gathering data from three or four recent projects is enough to move beyond a rough estimate — the difference between the perceived time and the actual time is usually the first surprise.

The hourly cost includes the profiles of the professionals involved. A cost estimator, a quantity surveyor, and a manager do not cost the same per hour.

The number of cases not closed over twelve months. Not lost business in the commercial sense: cases produced, submitted and left without follow-up.

The product of the three. That's the annual cost of producing incomplete responses.

Three points to note. This result isn't a loss to be eliminated: responding to unsuccessful bids is part of the job, and a 100% success rate would primarily mean that the pricing is either too high or too low. The value of the calculation lies in having a rough estimate to help make informed decisions. Furthermore, this figure is only meaningful when based on the company's specific data: industry averages are meaningless when an electrical engineering project and a finishing trades package don't have the same scope of studies. Finally, the relevant comparison isn't "this cost versus zero," but rather "this cost versus what it would be if it were reduced by a third."

The two costs that the calculation does not capture

The first issue is that of applications that go unanswered. When three tender documents arrive in the same week and the research team can only process two, the decision is often made without a cost-benefit analysis, due to a lack of time. This cost doesn't appear in any table: the company will never know what it would have obtained from the rejected application.

The second is structural. What allows for accurate cost estimation—proven ratios, trusted suppliers, pitfalls identified in a technical specification, the method for estimating a poorly defined position—generally exists nowhere but in the memory of one or two people. Peter Drucker stated this as early as 1954: when knowledge resides only in people's heads, the company is fragile. A three-month sick leave in the cost estimating department, and it's the ability to respond that stops, not just one person. In a sector where recruiting skilled professionals remains challenging, this concentration of expertise is an exploitation risk, not a human resources issue.

What the tools change — and what they don't change

Artificial intelligence solutions applied to calls for tenders have proliferated over the past two years. It is important to be precise about their scope.

What they can do today, and quite well: sift through a large tender document, extract the components and constraints, help decide between Go/No-Go decisions, and produce a first draft of the technical report. The time saved in reading and sorting is significant.

What a generic tool can't do: provide a quote. Estimating a cost item requires an up-to-date price database, company-specific ratios, knowledge of its suppliers, and its measurement practices. A system that doesn't have this information doesn't produce a quote: it produces a market average, which is exactly what a cost estimator doesn't want.

Two practical consequences for anyone evaluating a solution. Explicitly ask on what data the proposed estimate is based—the company's own data or an external benchmark. And verify what the company retains the day it stops paying: in-house expertise or a loss of access.

Fundamentally, the question isn't whether AI will replace the human programmer. Pricing decisions and margin risk assessment remain, and must remain, human. The question is whether companies are still willing to allow the mechanical aspects of the work—re-entering, copying, and proofreading 200 pages at 23 p.m.—to continue consuming their most skilled research resources.

Where to start

Without a budget or tools: time the actual time spent on the next three projects, step by step. This is the only worthwhile starting point. Most managers who do this exercise discover that the time spent on each step doesn't match what they imagined—and it's this distribution, more than the total, that indicates where to take action.

Opinion piece by Vincenzo Ounas, consultant in cost estimation automation for the construction industry (CapDurence Project, Hauts-de-France). Formerly a project manager, he has prepared cost estimates for high-voltage substation and railway projects at Eiffage Énergie Systèmes and Bouygues Energies & Services ( LinkedIn ).

Sources

OECP, Public Procurement Figures 2024, March 2026 — PDF, economie.gouv.fr (223.383 contracts, €233,3 billion, share of SMEs). Summary: Seban Avocats, April 16, 2026

FFB, 2025 Review and 2026 Forecast, December 2025 — ffbatiment.fr · full report (PDF) (+1,8% in 2026, new housing +9,5%, non-residential +0,5%, -4% in 2025)

FNTP, 2026 Forecast: What are the prospects for public works activity?, November 28, 2025 — fntp.fr (-1,9% in value, or -3,2% in volume)

Altares, Business Failures and Safeguards in France – Q2 2026, July 16, 2026 — altares.com (4.148 failures in the construction sector, -0,8%)

Peter F. Drucker, The Practice of Management, Harper & Row, 1954