First, establish comparable work
The default sample uses the same state, agency, exact definition, compatible specification edition and compatible unit. The period is the 36 months ending on the dataset snapshot date. Similar descriptions are candidates for review, never evidence of equivalence. No cross-state crosswalk is asserted in this release.
One contract, one final weight
For each contract, take the median of the eligible bidder unit prices. Then take the median, 25th percentile and 75th percentile of those contract medians. A contract with ten bidders receives the same final weight as one with two. We display both the number of contracts and the number of underlying bids.
A reproducible percentile rule
We use Hyndman–Fan type 7. Sort n values, find index (n − 1) × p, and linearly interpolate between its neighboring values. For contract medians 10, 20, 30, 40, 50, P25 is 20, the median is 30 and P75 is 40. This empirical spread is not a confidence interval or a prediction.
The minimum sample is a policy
Fewer than five comparable contracts means Insufficient comparable contracts and no line estimate. Five to nine is a limited sample. These are editorial thresholds, not guarantees of statistical reliability. Older observations remain visible even when they fall outside the default period.
Which observations are excluded
Zero or negative prices and quantities, lump sums, incompatible or unknown units, unproved specification compatibility, adjusted or unknown price bases and confirmed data-quality problems do not enter a unit-price benchmark. Technical duplicates are removed in normalization. Different bidders are retained. Repeated contract lines are excluded unless their scope compatibility is documented. Every reason remains visible.
Extremes are flags, not verdicts
Values outside the contract-distribution outer fences, P25 − 3 × IQR and P75 + 3 × IQR, are flagged for review. They are not dropped merely because they are unusual. Manual observation exclusions are listed in the result and exports.
Quantity, units and money
The optional quantity range is 0.5q–2q; it is a product setting. Users can change the period, region and quantity bounds. Only fixed, dimension-compatible unit conversions are supported. Short tons and metric tonnes remain distinct units. We do not infer density or thickness. Arithmetic uses a decimal library with 32 significant digits and half-up rounding; monetary displays and exported line amounts round to cents.
Reading the worksheet
A historical line scenario is quantity × the historical median unit price. P25 and P75 scenarios use the same quantity. User price deviation is (user unit price − historical median) / historical median × 100%, after compatible unit conversion. The Matched-items subtotal includes only matched rows with sufficient data. There is no automatic tax, profit, overhead or mobilization allowance.
Nominal bids and source limitations
Prices are nominal historical USD. We do not apply inflation adjustment. Caltrans adjusted prices are preserved separately and excluded from the default benchmark. Oregon’s calculated unit price is the source extended bid amount divided by quantity; it is not inflation normalization. Neither source establishes final paid amounts in this collection.