Define the metric
Examples include cost per unit of capacity, square metre, user, transaction, kilometre or installed component. State whether the denominator represents design capacity, average use or delivered output.
Normalize the data
- Currency and price date
- Location and labour market
- Included and excluded scope
- Size, complexity and technical standard
- Delivery method and schedule constraints
- Taxes, owner costs, escalation and contingency
Use distributions, not one magic number
A median, quartile range and explanation of outliers is usually more useful than a single average. Small samples or confidential source data should be described honestly.
Benchmark the estimate and the process
Compare not only the final unit cost, but also labour productivity, material quantities, design maturity, change rate and contingency use. This helps distinguish price differences from execution differences.
Red flag: a low benchmark may omit owner costs, support systems or commissioning that your project requires.