/platform/data-tables) is the workspace for comparable metric extraction across a watchlist. You write one prompt describing the metrics you want, and the platform runs it against every entity, normalising the results into a sortable, filterable matrix.
When to use Data Tables vs other surfaces
The difference from Deep Dive is the shape of the interaction: Deep Dive is a conversation about a single obligor where you ask follow-ups and refine the answer; Data Tables is a one-shot extraction across the entire watchlist.
The workspace
There’s one workspace per prompt. The sidebar lists every prompt you’ve authored. The main area has three sections:- Prompt panel: the prompt body. Editable inline.
-
Matrix: rows (typically entities) × columns (typically fiscal years or another comparable dimension). Every value is traceable: click a cell for its reference, source document, pages, and notes, or click an entity name to open a panel with two tabs:
- Documents: the source filings used for that entity.
- Notes: caveats, ambiguities, methodology, and restatements flagged during extraction, with links to the cited pages.
- Views and filters: switch between the three views and filter by entity, status, value type, period, and audit status.
Running a table
Run opens a modal where you choose which of the watchlist’s CUSIPs to extract. Start small, then extend the same table:1
Run on a few CUSIPs
Click Run and pick a handful of representative CUSIPs. Extraction covers
only your selection, so trying a prompt out is cheap.
2
Tune the prompt
Adjust the prompt body until the cells look right. Re-running an edited
prompt replaces the table with a fresh generation.
3
Extend to the rest of the watchlist
Once the prompt is dialled in, click Run again and select the remaining
CUSIPs. Already-extracted entities are locked and keep their cells; only the
newly selected ones are extracted, and the matrix fills in as they come back.
Options
Alongside the prompt, a few choices control what the extraction runs over:Views
The same extracted cells can be sliced three ways. Switch between them from the toolbar:Tips
- Be explicit about units and time bases. “Revenue in USD millions” and “Fiscal year ending June 30” beat “operating revenue” for sortable comparability.
- Pin sector idiosyncrasies in the prompt. For example, exclude entrance fees from DSCR for charter schools, or specify GASB vs FASB.
- Start on a small selection. Run the prompt on a few CUSIPs first, then tighten definitions, units, and output shape before extending it to the rest of the watchlist.
See also
- Writing effective prompts: how to write prompts that extract clean, comparable values.
- Watchlists: the groups of entities a Data Table runs across.
- Surveillance: the event feed for new disclosures on those same watchlists.