Data & sources
Every input VPM uses, where it comes from, and how it is licensed — plus a clear line on what VPM sells and what it never redistributes.
The inputs
VPM combines five kinds of data: where vessels are now, where they have been, where ports are, how sea routes actually run, and what each vessel's machinery is. Everything the product outputs is built from those.
Live position data
- AISStream.io — a free AIS WebSocket feed; the primary live source.
- Finnish Digitraffic — the Finnish transport agency's open AIS, covering the Gulf of Finland and Bothnian traffic.
- Data Docked — a paid terrestrial AIS feed held as an outage backup, polled on a 12-hour cycle. Used only when the free feeds are down.
Terrestrial AIS is the right fit for the Baltic — the coastline is densely covered by shore stations, so satellite AIS would be paying for reach that isn't needed.
Ports and sea routes
- Port coordinates — UN/LOCODE, via a community-corrected edition of the code list. Open data.
- Baltic shipping-lane density — HELCOM's 2024 AIS shipping-density grid, ~1 km resolution. Public.
- North Sea, Channel and Biscay density — EMODnet Human Activities vessel-density, published under CC BY 4.0.
The density grids are what let VPM measure a route the way a ship sails it — following known lanes, through the Kiel Canal when that is shorter — instead of drawing a straight line across land.
Vessel particulars
Engine power, service speed, fuel type and grade, tank capacity, build year, tonnage — collected by hand, per vessel, from classification-society machinery records and owners' published fleet specifications. This is slow, deliberate work; it is also why the fuel model has something real to stand on.
Generic "fuel type" fields in commercial databases are unreliable — often blank, or a catch-all like "Marine Diesel" that doesn't say whether a ship actually burns MGO or heavy fuel. Owners' fleet specs are the source that holds up.
Contact & technical enrichment
To fill gaps in a vessel's profile — operator, owner, email, phone, bunker contact — VPM queries a large language model (Google Gemini) with live web-search grounding, so answers come with cited sources rather than from model memory.
- It returns candidates, never facts. Every value is presented for a human to confirm before it is saved.
- Nothing is written automatically. This is a deliberate choice — against both hallucinated details and storing personal data without a check.
The calibration anchor
The fuel model is anchored to a single real, reported consumption figure from one vessel's technical file — a known daily burn, at a known speed, laden. Every other vessel's burn rate is scaled from its own machinery and speed to that anchor. When a second reported figure turns up, the anchor is re-checked against it.
Public, licensed, and ours
| Layer | Status |
|---|---|
| UN/LOCODE, HELCOM density, EMODnet density, Digitraffic AIS | Public / open |
| AISStream feed | Free API |
| Data Docked feed (backup), vessel specs & history, LLM enrichment | Paid |
| Movement models, calibrated fuel model, route graph, port dictionary, accuracy scoring, the Port View | VPM's own |
The last row is the product. The rest is raw material.
What VPM does not resell
VPM sells decisions and estimates — a predicted port, a fuel need, a cost, an arrival window. It does not redistribute raw AIS positions or any provider's underlying dataset. Where VPM sources data under licence, it favours terms that explicitly permit derived, value-added products; where a provider's terms are narrower, that source is used only internally and never in customer-facing output.