The Same Container, Seven Sources, 3,250 kg of Disagreement

We took the sources that actually get read: the pages that rank on the first page for container dimension queries, and the pages that Google’s AI Overview and ChatGPT cite when asked. For each we recorded internal dimensions, tare weight, payload and cubic capacity exactly as published, converting units but never rounding to make a figure fit.

Where a publisher gives a range we kept both bounds. Where a figure contradicts something else on the same page, we recorded the figure and noted the contradiction rather than choosing which half to believe.

Two things we deliberately did not do. We did not rank sources by quality — this is a measurement of spread, not a league table. And we did not treat our own figures as the answer: contsize appears as one source among seven, and the spread is computed with our number inside it, not against it.

Carrier specification sheets from Maersk and Hapag-Lloyd were not machine-readable at the time of collection — both refuse automated requests — so this first edition compares the secondary sources that the search and AI layer actually consumes. Adding carrier primaries is the next edition.

Read the spread column first. Dimensional figures agree closely — internal height of a 40 ft high cube spans just 10 mm across five sources, which is ordinary manufacturing variation and nothing to worry about.

Weights are a different story. Payload is the number that decides whether your cargo is legal, and it is the number sources disagree about most: 3,250 kg on a 20 ft box, 3,000 kg on a 40 ft high cube. Anyone planning a load from the wrong row is three tonnes out before they start.

Three mechanisms explain almost all of it. The first is honest: payload equals max gross minus tare, and tare genuinely varies by manufacturer and build year, so a 150 kg difference in tare moves payload by the same amount. That accounts for the small spreads.

The second is a ratings mix-up. Most boxes are plated for a max gross of 30,480 kg, but heavy-tested ones are rated to 32,500 kg — and ISO 668 has allowed up to 36,000 kg since 2016, so the rating is the manufacturer’s choice, not a single figure in the standard. Quote a 32,500 kg rating against a standard tare and you publish a payload that looks generous by two tonnes and is not available on most equipment. One source in our table states “up to 32 tonnes with carrier approval”, which exceeds 30,480 kg once tare is added back.

The third is simple copying. A figure enters one popular page, gets quoted by the next site without a source, and by the tenth repetition nobody can trace where it came from. That is how a 25,000 kg payload for a 20 ft container — over three tonnes below the ISO arithmetic — ends up on a global freight forwarder’s own site.

One pattern deserves its own paragraph because it is the most consequential error we found. A 40 ft standard container holds 67.7 m³ and stands 2,591 mm tall externally; a 40 ft high cube holds about 76 m³ and stands 2,896 mm. They are different boxes that cost different money.

One page in our comparison — ranked on the first page in Mexico and cited by Google’s AI Overview for Spanish dimension queries — publishes 2.90 m external height, 2.69 m internal height and roughly 76 m³ under the heading “contenedor de 40 pies” with no high-cube qualifier. Those are high-cube figures. A reader planning around them will book the wrong equipment, and an assistant summarising that page will repeat the mistake.

We are not naming this to score a point. We are naming it because the error is currently propagating into AI answers, and the only way that stops is if somebody writes down what the correct split is and shows their working.

If you are planning a load, the operative number is not on any website: it is on the CSC plate bolted to your specific container, which carries that unit’s real tare and max gross. Every figure published anywhere — ours included — is a typical value for a class of equipment.

If you need a planning figure before the box arrives, use the ISO arithmetic rather than a quoted payload: max gross 30,480 kg for standard ratings, minus the tare band for that type, and leave margin. Our per-type figures are in the container weight guide.

And if you are writing about containers, cite the standard and the date. Half the spread in this table exists because somebody published a number without saying where it came from.

The comparison is published as an open dataset — spec-variance.json and spec-variance.csv — free to reuse with attribution. Each row carries the publisher, the figure, the page it came from and the date it was read, so the measurement can be repeated or contradicted.

If we have misread your page, or you have corrected a figure since we read it, tell us and we will re-read and re-publish with a new date. Our method is on how we verify data, and corrections are logged on the about page.