Minimizing Bias to Accurately Forecast True Costs for Construction Projects

Forecasting inaccuracies in preconstruction compound into cost and schedule overruns. Behavioral drivers—planning fallacy, optimism bias, and heuristics—distort early estimates. OntaBuild adopts an RCF + unpacking approach that compares projects against archetype libraries and analyzes each scope as a distribution rather than a single point estimate. Across eight completed projects, our preconstruction estimates were within 0.52% below actual direct costs (avg.) and only 1 of 8 (12.5%) experienced a cost overrun—dramatically outperforming industry baselines where 73.66% of building projects overrun and budget/schedule overruns average 27% and 38%, respectively.

Results reflect projects completed using OntaBuild's RCF + unpacking workflow.

• Across 8 completed projects, mean precon estimate was 0.52% below actual direct costs.
• Only one project overran (12.5% of sample), driven by a landlord-mandated vendor with a 7× price anomaly flagged by our method.
• Observed linkage between high-variance scopes and later change orders (further study in progress).

Assuming the 73.66% building overrun rate from the Flyvbjerg dataset (n=186) represents the population baseline, a test of proportions indicates OntaBuild's observed 12.5% overrun rate (1 of 8) differs significantly at p<0.001 (95% confidence). While early-stage and underpowered, the effect size is large and directionally clear.

Note: We will continue adding larger $20M+ projects to strengthen power and extend generalizability.

• Variance-first views expose design ambiguities and estimator omissions before contract award.
• Targeted clarifications shrink distributions, bringing true cost into clearer view.
• We suspect high bid variance correlates with increased change orders; ongoing research aims to quantify this signal.
• Our approach helps "cut the tail"—reducing the likelihood of fat-tailed (≥50% overrun) outcomes reported in industry data.

OntaBuild's RCF + unpacking workflow demonstrably reduces bias in preconstruction estimates and improves downstream delivery certainty. As our dataset grows across larger programs, we expect even stronger evidence that platform-driven forecasting can meaningfully lower the probability and magnitude of overruns, enabling owners and GCs to make faster, better capital decisions.

• Flyvbjerg & Gardner (2023). How Big Things Get Done.
• Flyvbjerg et al. (2025). The Uniqueness of IT Cost Risk.
• McKinsey (2017). Reinventing Construction.
• Rivera et al. (2016). Identifying the Global Performance of the Construction Industry.
• Sassano (2025). The holistic view in forecasting.