Enter a rate you were quoted or paid for haulage, rail, barge, port charges or sea freight. You get a verdict in seconds against our open estimate and, as rates come in, against what others paid.
Our estimate for this route, worked out month by month from published fuel, ship-hire and cost data, .
Rates other shippers added for each part of this route, adjusted to this route's distance so different trips compare fairly. A median appears once five or more rates are in.
Each part of a route becomes a live market benchmark once five rates are in. The first person to add one is credited for starting it. Pick any cell to check or add a rate there.
For a route that isn't listed, or to record tonnage and contract terms. Rates are only ever shown without names, as part of a group of five or more.
All routes, newest first. Individual prices are never shown, only medians of five or more.
| Part | Route | Distance | Month | Basis |
|---|
The full cost model with every formula, how each part has been tested against real prices, and where the data comes from. For analysts, data partners and anyone who wants to check our working.
The verdict compares your rate with this model. Every input can be changed and every formula is shown, and changes here update the verdict above.
How well each part of the model matches real prices it was not fitted to, and every number it relies on with its source. A part is cleared to publish as a benchmark only when it passes the gate: average error 15% or less, bias 5% or less, at least 8 independent checks.
Estimates, not contributed rates. Each part carries its test status: Pass means that on prices the model was not fitted to, it is within the publish gate (average error 15% or less, bias 5% or less, at least 8 independent checks). Ocean: 31 out-of-sample checks (ISM, USDA and 12 months of AHDB Odessa→Rotterdam freight), average error 13.0%, bias −5.5%, just outside the gate. Coaster: fitted to AHDB Hamburg→UK freight and tested on 2023 (14.2%, bias −13.6%). UK road: passes the 2013 haulage-survey hold-out (5.0%) but fails the AHDB delivered-minus-ex-farm test for 2021–26, where spreads fell while modelled costs rose; recent haulage may be over-priced. The full list of assumptions is below.
V verified from a cited source · D derived from cited figures · C calibrated to observed rates and then tested out-of-sample · E estimate, untested and first in line for replacement.
The model is refreshed every Monday from published data, re-tested, and only updated here if every check passes.
The licence model this prototype is built around.
Every formula is on the page and the code is intended for an MIT licence. Anyone can audit, copy or extend any part of it, which is what makes the number trusted.
Anyone gets a verdict against the model. Adding a real rate unlocks market medians and percentiles for 90 days, so the data stays fresh.
Rates are never shown individually or with names. A median appears only when five or more rates sit behind it.
Anonymised rates are free for non-commercial use under a share-alike database licence. Feeding them into a paid product, index or API needs a commercial licence.