The Data Desk · Downloads

Take the numbers with you.

The dataset behind the site, regenerated with every build: the tests on the Docket, the results, who is on each side, the companies and people, and the tape. CSV for a spreadsheet, JSON for code.

12 files · built October 8, 2026

Back to the Data Desk

  1. tests

    46 rows · 16 columns

    Every dated test on the Docket: one row per watchlist item, with its ISO date and precision, status, the source of the date and how it resolves.

    CSV · 13.2 KB JSON · 26.9 KB

    Columns: id, part, story_slug, story_headline, index, when, date, precision, status, result_id, event, source_kind, source_ref, source_url, resolver_kind, resolver_rule

    Source. DeepStack’s stories and content/tests.json; dates come from the published FOMC calendar, the Financial Modeling Prep economic and earnings calendars, and the stories themselves, named per row in source_ref.

    License. CC BY 4.0 for DeepStack’s own fields (attribution: DeepStack, deepstack.ltd). Third-party figures keep their owners’ terms, named in the row that carries them. Analysis and opinion, not investment advice.

  2. results

    0 rows · 11 columns

    How each test came in: verdict, headline, the paragraphs, the evidence with its source and date, and what it did to the Call. Published only after a human review.

    CSV · 95 B JSON · 3 B

    Columns: id, test_id, part, story_slug, resolved_on, verdict, headline, body, evidence, call_update, approved_at

    Source. DeepStack results; every evidence row names its primary source (FOMC statements, FRED, BLS, company filings). No Financial Modeling Prep or Polygon figure is used as evidence.

    License. CC BY 4.0 for DeepStack’s own fields (attribution: DeepStack, deepstack.ltd). Third-party figures keep their owners’ terms, named in the row that carries them. Analysis and opinion, not investment advice.

  3. sides

    198 rows · 13 columns

    Who is on each side of each story’s question: one row per voice, with the side, the paraphrased position, the date and a link where one was verified.

    CSV · 87.5 KB JSON · 146.9 KB

    Columns: part, story_slug, question_id, question, side_id, side_label, voice, entity_slug, date, claim, url, access, link_verified

    Source. Paraphrases of public writing, attributed by name; the authors own their words and are quoted within fair use. DeepStack’s structure, side labels and ids are its own.

    License. CC BY 4.0 for DeepStack’s own fields (attribution: DeepStack, deepstack.ltd). Third-party figures keep their owners’ terms, named in the row that carries them. Analysis and opinion, not investment advice.

  4. companies

    33 rows · 9 columns

    The companies in The Buildout: ticker, the parts each appears in, the one-paragraph summary assembled from the stories, and the counts of reported figures, mentions and positions.

    CSV · 9.0 KB JSON · 120.3 KB

    Columns: slug, name, ticker, parts, summary, reported_figures, mentions, positions, last_updated

    Source. Assembled from DeepStack’s stories. Reported figures reference the sources named in each story; no market quote is included.

    License. CC BY 4.0 for DeepStack’s own fields (attribution: DeepStack, deepstack.ltd). Third-party figures keep their owners’ terms, named in the row that carries them. Analysis and opinion, not investment advice.

  5. people

    64 rows · 9 columns

    The people whose arguments the stories weigh: role, the parts each appears in, the summary, and the counts of mentions and attributed positions. No score, ranking or accuracy, for anyone.

    CSV · 18.1 KB JSON · 178.2 KB

    Columns: slug, name, role, parts, summary, reported_figures, mentions, positions, last_updated

    Source. Assembled from DeepStack’s stories; what a person argued is reported and attributed, never rated.

    License. CC BY 4.0 for DeepStack’s own fields (attribution: DeepStack, deepstack.ltd). Third-party figures keep their owners’ terms, named in the row that carries them. Analysis and opinion, not investment advice.

  6. tape

    40 rows · 7 columns

    The tape: every figure reported in The Buildout, one row each, with the story it comes from and the source and date in context.

    CSV · 6.4 KB JSON · 11.6 KB

    Columns: part, story_slug, story_headline, published, value, label, context

    Source. Figures as reported in the stories, each naming its source in the context column (FactSet, CNBC, company filings, the authors cited). Third-party figures keep their owners’ terms; DeepStack’s selection and labels are its own.

    License. CC BY 4.0 for DeepStack’s own fields (attribution: DeepStack, deepstack.ltd). Third-party figures keep their owners’ terms, named in the row that carries them. Analysis and opinion, not investment advice.

How to cite. DeepStack, “The Buildout dataset,” deepstack.ltd/data/download/, retrieved on the date of download. The files change with every build; keep the build date above with any copy. No market quote from a data vendor is included. Analysis and opinion, not investment advice.