MethodologyWhat counts as a source, and when we refuse to publish a number
How every figure on this site is produced, and when we refuse to produce one
Every number here belongs to one of four classes, and the class decides whether it can be published at all. A figure read from a primary record can be published with its reading date. A figure from a secondary source is published with the publisher named. Our own count is published with the corpus and the threshold stated. And a figure we do not hold is named as missing rather than estimated.
The four classes of figure
| Class | What it means | How it is published |
|---|---|---|
| Primary | Read from the record itself — a register, a tariff, a fee schedule | With the reading date beside it |
| Secondary | Published by a named third party — an index, an industry table | With the publisher and the period named |
| Our own count | Measured by SeaRates across a stated corpus | With the corpus size, the threshold and the sweep date |
| Not held | Needed for a claim, and we do not have it | Named as missing — never estimated to fill the gap |
A figure that cannot be placed in one of the first three does not go out. That is why several pages here publish the rows a ranking would need and mark which ones are absent, instead of ranking on adjectives.
How the corpus was read
Before writing, SeaRates captured the full text of pages ranking in US search across 22 question clusters and counted what they contain. The method is stated so the count can be repeated: full page text captured alongside the search results, pages under 200 words excluded, and the threshold for "publishes a method" set at 120 words describing how a result was produced.
- 2,001 pages read in full across 22 clusters, 07/25-26/2026.
- 245 pages measured across the ranking clusters; 2 show a licence number.
- 4,967 unique People Also Ask questions collected; 2,410 survived filtering.
Counts of this kind are our own and are labelled as such. They measure what the market publishes, not whether it is true.
When a figure is refused
Three refusals recur, and each is visible on the site:
- No average where none exists. Import duty depends on the HTSUS line, the country of origin and the measures in force. The landed-cost calculator takes a duty rate as an input rather than inventing one.
- No median without an array. SeaRates holds no quote array, so no page here publishes a median rate. Published vendor bands are shown as bands and labelled as bands.
- No sum of two registers. The FMC forwarder and NVOCC lists overlap — 2,588 organizations hold both roles — so the two counters are published separately and never added.
Dates, and why every figure carries one
The sources behind this site move. The FMC register changed by nine records in the two days between 07/26/2026 and 07/28/2026. Ocean spot rates reprice weekly. A figure from either without its reading date is not a fact about the source, and pages here are built so the date travels with the number rather than sitting in a footer.
Where two sources disagree, both are published with attribution rather than one being chosen. Schedule reliability for mid-2026 is the standing example: Sea-Intelligence and a Kuehne+Nagel summary give different figures for different months, and both appear.