Cost benchmarking

Benchmarking answers the question an estimate cannot answer about itself: is this normal? Metrics are chosen at a comparable level — cost per square metre, per tonne of steel, per installed megawatt, or element ratios — and the reference data is normalised before comparison: indexed to a common price date, adjusted by location factors, and screened for scope differences such as who carried the owner's costs or whether the site infrastructure existed.

Done well, benchmarking is a diagnostic, not a verdict. An estimate 30% above the benchmark set is not wrong; it is carrying something the reference projects did not, and the useful output is finding out what. Reviewers and lenders use it exactly this way, as a structured request for explanation.

Done badly, it is the most efficient way to launder a false number. Unnormalised comparisons — different years, different countries, different scope boundaries — can justify almost any figure, and a benchmark quoted without its basis deserves the same suspicion as an estimate quoted without one.