Methodology
Sub-score weights, data sources, freshness and confidence - published in full so a credit committee, a marine underwriter or an academic can audit the working.
The composite vessel risk score is a weighted average of nine domain scores. Weights and inputs below; full inputs and code documented at /docs/methodology in the portal.
Each field on every screen comes back with a provenance block - the source it came from, when we fetched it, and the confidence the platform assigned. A real response shape:
{
"field": "flag_state",
"value": "Marshall Islands",
"_provenance": {
"source": "ihs_sea_web",
"fetched_at": "2026-05-13T08:15:00Z",
"fetch_method": "api",
"confidence": 0.99,
"freshness_days": 1
}
}The same block appears on PDF and Excel exports. When the value is uncertain, the confidence drops and the consumer sees it - no silent defaults.
Each data domain has a target refresh cadence. The actual freshness is published live on the platform's data-quality dashboard. Typical SLAs:
| Domain | Refresh target | SLA (P0 entities) |
|---|---|---|
| Sanctions | Hourly | 4 hours max |
| PSC inspections | Daily | 24 hours max |
| Ownership | Weekly | 7 days max |
| AIS positions | Continuous | 15 minutes max |
| News and media | 15 minutes | 1 hour max |
| ESG / CII | Quarterly | 90 days max |
No anonymised aggregators. No "proprietary feed". Every input on the platform is either a public regulator, a licensed commercial source or our own measurement.
Complete source registry with licence terms is at /accuracy.
The free tier gives you one vessel screen a month. The methodology page in the portal links straight to the underlying SQL and Python.