How the record is made.

Each source contributes the things it does best. We never invent values — every field on a satellite page traces back to one of these:

• CelesTrak — real-time orbital elements (the numbers that describe where an object is and how it moves), plus NORAD catalog numbers and object names. This is our backbone for “what is in orbit right now.”
• GCAT (Jonathan McDowell) — the authoritative launch catalog: launch and decay dates, object classification (payload, rocket body, debris), operator, country, and orbital regime.
• Wikidata — community-maintained facts that the other two don’t carry: operator details, country of operator, launch mass, and mission purpose.

We deliberately do not use Space-Track.org data, whose terms restrict redistribution. See Sources for every source, its license, and its attribution requirements.

Data is refreshed automatically on fixed schedules, not by hand:

• CelesTrak — fetched daily at 06:00 UTC by an open-source GitHub Actions workflow. Orbital data changes constantly, so it gets the most frequent refresh.
• GCAT — downloaded approximately weekly. The launch catalog changes slowly.
• Wikidata — queried via SPARQL on a slower cadence, since the fields we pull (operator, mass, purpose) rarely change once set.

The ingestion scripts and the workflow definition are public, so anyone can audit exactly what we fetch and when.

The hard part of a project like this is deciding when two records from two different sources describe the same physical object. We do it in three steps.

Every stored value (we call them “claims”) carries a confidence score from 0.0 to 1.0 reflecting how much we trust its source for that kind of field:

Displaying the score lets you decide for yourself how much weight to give any single value rather than trusting a black box.

We would rather state our weak spots plainly than have you discover them:

• ·Uneven Wikidata coverage. Large constellations are sparsely documented — many individual Starlink satellites have no Wikidata entry, so operator/mass/purpose may be missing for them.
• ·Under-documented status. Non-operational status is poorly recorded for small operators, and some objects can show an “unknown” operational status when no source supplies one. About 4% of the catalogue (roughly 620 objects) currently has an unknown status pending source data.
• ·Archive depth. Historical orbital completeness depends on CelesTrak’s archive. Some objects out of CelesTrak’s active set (for example, very high orbits like JWST at L2, or long-decayed objects like Vanguard 1) may have limited or no current orbital data.
• ·Orbital data is approximate. Our elements are suitable for browsing and research, not for operational tracking, collision avoidance, or maneuver planning.
• ·Object-type coverage is payload-only. Our satellite records come from CelesTrak's active feed, which is almost entirely operational payloads. GCAT and Wikidata enrich those existing rows but never create new ones, so rocket bodies, debris, and long-decayed objects are barely represented: under 10 records total, despite GCAT's full catalog covering tens of thousands of them.

Publishing an honest error rate matters more to us than claiming perfection. We measure accuracy two complementary ways, and we publish updated results periodically.

The spot-check is regenerated by a committed, read-only script (scripts/spot-check.ts), and all validation logs are committed to the public repository so the checks are auditable. Errors found during validation are corrected before they reach the site. We re-run validation on each data refresh and update these figures here.

Found something wrong? We want to know. Every record is open to correction, and reports are reviewed privately by our team — we read every one.

Use the “Report a correction” control on any satellite detail page, or the form below. Please include the NORAD catalog number, the field in question, and a primary source for the correct value where possible — that lets us verify and fix it quickly.