How AI cargo–vessel matching works
Matching a cargo to a vessel looks simple until you try to automate it. A useful match has to understand what the cargo needs, what the vessel can do, and where the two genuinely line up. Here's how AI-supported matching gets there — and why the highest score isn't automatically the fixture to chase.
It starts with structured data
Enquiries and positions arrive as free text — in email bodies, forwarded threads and attachments. Before anything can be matched, that text has to become structured cargo and vessel records: commodity, quantity and tolerance, load and discharge ports, laycan; and on the vessel side, type, capacity, gear, draft and open position. Comparing structured records is the difference between matching and guessing.
Compatibility is more than keywords
A real fit depends on hard constraints that a keyword search never sees:
- Capacity and stowage against the actual stem
- Vessel type suited to the commodity and trade
- Gear where the terminal has no shore cranes
- Port and draft limits at both ends
- Laycan feasibility from the vessel's open position
Missing information is a first-class result
The most dangerous input is the one that isn't there. When a required field is absent or ambiguous, it should be surfaced — not quietly assumed — so a near-fit built on a guess never masquerades as a strong match.
Ranking with the reasons shown
A score on its own is not an answer. Useful ranking explains why an option fits, what could add cost or delay, and what still needs checking. That lets you compare the trade-offs across a shortlist rather than trusting a number.
Why the top score isn't always the best fixture
The highest-ranked option can carry a long ballast leg, a tight laycan or a gear question. The best fixture is the one with the strongest overall picture — suitability, voyage economics and risk together — which is exactly what an explainable shortlist is for.
A worked example
Picture a 25,000 mt grain stem loading at a geared-berth port. A gearless panamax quotes the lowest rate — but it can't work the berth, and re-positioning cranes or shifting berth eats the saving. A geared handysize a little further out quotes higher, yet ballasts a short leg and fits the terminal. On rate alone the panamax 'wins'; on the whole voyage, the handysize is the better fixture. Matching that shows its reasoning makes that visible before an offer goes out.
What structured matching changes
When both sides are structured and the constraints are explicit, the trade-offs stop being tribal knowledge held in one broker's head. The reasons a match works — or doesn't — travel with it, so anyone looking at the shortlist can see why the order is what it is.
Key takeaways
- Structured data is the precondition for real matching, not a nice-to-have
- Compatibility is capacity, type, gear, port and laycan — not keyword overlap
- Missing information should be surfaced, never assumed
- The best fixture balances suitability, economics and risk — not the top score alone