Structured, provenance-carrying facts about tracked companies, extracted by enrichment from collected items: one fact per row, with its dimension, value, and as-of date.
One row per extracted fact. Deduped per company at write time, so a fact is added once and kept. source_url resolves through the originating item and is null on the rare fact ingested without one. Add ?company=<slug> to narrow to one registry company. This is a working corpus, not a redistribution channel. Downloading requires an access key.
| Column | Type | Definition |
|---|---|---|
| fact_id | text | Stable UUID of the fact. |
| company_slug | text | The company the fact is about; joins intel-companies. |
| company_name | text | Company name, denormalized for convenience. |
| dimension | enum | The Intel Desk's taxonomy code: strategy, products, tech_ai, financials, leadership, regulatory, ma_partnerships, brand, talent, or risk. |
| fact | longtext | The extracted fact statement. |
| value_text | text | A structured value for the fact, when one was extracted. |
| as_of | date | The date the fact is reported as of, when known. |
| source_url | text | URL of the item the fact was extracted from, when known. |
| item_id | text | The collected item the fact was extracted from; joins intel-items. |
| created_at | date | When the fact was extracted (YYYY-MM-DD). |
unfiltered
Heavy dataset: the preview is capped at 25 rows; download for the full corpus.
Preview needs an access key. Unlock once and it opens up here.