Construction's productivity crisis: why ML cost forecasting and off-site methods are converging

U.S. construction sector labor productivity has declined consistently since 1968, even as manufacturing, communications, and virtually every other major industry has moved sharply in the other direction. That single statistic, drawn from McKinsey Global Institute research cited by the U.S. Department of Energy, frames two significant developments now giving project owners and facility operators new tools to act on the problem.

A structural problem, not a cyclical one

The DOE's Building Technologies Office attributes the productivity gap to underinvestment in innovation, industry fragmentation, and a persistent shortage of skilled labor, all factors that McKinsey identified as structural rather than temporary. The practical consequence is higher upfront construction costs, which in turn cause building owners to bypass energy-efficient technologies even when those technologies reduce long-term operating expenses.

The scale of the retrofit challenge makes the productivity gap particularly acute. The United States has more than 125 million existing buildings, according to the DOE, and more than half were built before 1980, predating modern energy codes. Yet in leading jurisdictions, only 1.75% of homes and 2.2% of commercial real estate are retrofitted in any given year. At those rates, the building stock improves far too slowly to meaningfully shift national energy performance.

Off-site manufacturing and digitization as the supply-side response

The DOE's Advanced Building Construction (ABC) Initiative addresses the supply side of this equation. The program, led by the Building Technologies Office, funds research into building technologies that can be deployed quickly with minimal on-site construction time, covering off-site manufacturing, robotics, and digitization of building design and construction workflows. The initiative also coordinates stakeholders on workforce training, business models, and service delivery, recognizing that technology alone does not close a productivity gap shaped by industry fragmentation.

The National Institute of Building Sciences' Off-Site Construction Council has highlighted scheduling, price, quality, and safety as concrete productivity benefits of off-site approaches, according to the DOE. Venture capital and large technology companies have increased investment in startups working in this space, signaling that commercial interest is catching up with the research programs.

The demand-side problem: cost overruns before a shovel hits the ground

Even with better construction methods, project budgets frequently go wrong before operational improvements can deliver value. A peer-reviewed study published in Nature by researchers at Umm Al-Qura University in Makkah, Saudi Arabia, examined cost overrun patterns across building construction projects and found that 70% of the most significant contributing factors were internal project control deficiencies. Scope and design changes, inadequate documentation, and poor site management ranked among the primary causes, not commodity price swings or external market forces.

That finding matters operationally. It means the majority of cost overrun risk sits inside the project team's control, and therefore inside the scope of better forecasting and process discipline. The researchers surveyed and interviewed industry experts, assessed 49 pre-identified factors using the Relative Importance Weight method, and then built and compared seven distinct machine learning architectures to find the most accurate predictive approach.

A validated ML model with sub-3% error

The model that emerged as the strongest performer was a General Regression Neural Network. According to the Nature study, the GRNN achieved a mean absolute percentage error of less than 3% and a correlation coefficient of 0.95 against actual project outcomes. The researchers describe it as a decision-support system for expenditure optimization rather than an academic exercise, and the open-access publication makes the methodology available for teams evaluating similar tools.

For procurement directors and capital project managers, the combination of a high-accuracy forecasting model with a clear root-cause map of cost drivers offers a practical starting point for pre-project risk assessment. The study's framing around value preservation rather than loss prevention is also notable: the goal is maintaining project value through proactive control, not simply detecting overruns after the fact.

Where the two threads meet

The DOE initiative and the Nature study address different parts of the same problem. One targets the construction process itself, making it faster, more energy-efficient, and less dependent on on-site labor. The other targets the financial planning layer, giving project controllers a validated tool to forecast where budgets will break before ground is broken. Neither is sufficient on its own.

For enterprise operators managing large building portfolios, whether in corporate real estate, healthcare, higher education, or government facilities, the near-term question is how quickly these tools move from research programs into standard procurement and project management practice. The DOE is actively coordinating stakeholders to accelerate that transition; the Nature study's GRNN model is already published and citable. The next step is adoption.

What this means for your team

• Audit your retrofit pipeline against the 1.75% residential and 2.2% commercial annual retrofit rate benchmarks cited by the DOE to identify where your portfolio lags and where off-site construction methods could compress project timelines.
• Map your recent capital project cost overruns against the internal control deficiency categories identified in the Nature study (scope changes, documentation gaps, site management) to determine how much of your historical overrun exposure was controllable.
• Evaluate ML-based cost forecasting tools against the GRNN benchmark of sub-3% MAPE; any vendor claiming predictive accuracy should be able to demonstrate comparable validation methodology on real project data.
• Track the DOE Building Technologies Office's ABC Initiative for technology certifications, workforce training programs, and procurement frameworks that will define purchasing criteria for advanced construction methods over the next several years.