Every CEO interview, transcribed and verified.

Yes. Every transcript is split into speaker turns, and each turn carries the speaker's full name, a stable numeric ID that matches the executive's record in our database (empty for hosts and other guests), a timestamp into the source, and the cleaned text. You can read exactly who said what, and isolate the executive's own words from the interviewer's in one filter, in the web app or through the enhanced transcript field in the API.

Automated sourcing finds the appearance, a first model pass checks the speaker matches the source, a human reviewer confirms the executive is the one speaking and screens for personal or nonpublic information, then a second pass adds structure. Reposts of the same interview collapse to one record.

9,000+ executives on the record and 34,000+ tracked, across 4,700+ companies: 100% of the S&P 500, 100% of the Nasdaq-100, European and Chinese large caps, the Fed and Treasury, and 100+ AI labs and private companies. Browse the company list.

Yes. Every record carries three timestamps: the appearance date, when the executive actually spoke; the publish date, when the source went public; and an update time, when we last touched the record, alongside the time it entered our database. The corpus is append-only. New appearances are added daily and nothing is backdated, so a backtest only sees what was public on that day. Filter on publish date to reproduce exactly what was knowable at the time, and on update time to pick up incremental changes.

Yes. Every record carries the company ticker, SEC CIK, and legal name, the executive's full name and title, and stable numeric IDs for every executive and company. Those are the keys used to join to ExecuComp and Compustat (ticker or CIK to GVKEY), CRSP (ticker to PERMNO), BoardEx, Bloomberg, FactSet, and 13F holdings. The API returns all of it as JSON, and Substack posts carry the publication and author.

Yes. Every record is a public, on-the-record appearance, so the corpus contains no material nonpublic information and no personal data by construction, and a human reviewer screens each record for both before it ships. We have been through compliance and vendor reviews with hedge funds and with the data platforms and suppliers that redistribute our data, and we provide a DDQ, security questionnaire, and vendor packet on request. Read the data compliance page.

Earnings calls are scripted and lawyer-reviewed, four times a year. This corpus is the unscripted conversation in between: the podcast a week after the call where a CEO walks back the guidance, the conference panel where a CFO says what the prepared remarks left out, the TV hit where the tone changes before the numbers do. We capture those moments within a day of the appearance, attribute every turn to the speaker, and timestamp each claim to the source, so you can line up what an executive said on the call against what they said off script and see the gap in structured form. Use both.

Yes. The self-serve Data API is $499 a month, includes the full web platform, and lets you pull 100,000 transcripts and 200,000+ quotes a month, enough for a multi-year study of your watch-list companies. We recommend it for academics and individual prosumers. Companies and investment teams get an Enterprise license with the full corpus, a daily stream, bulk exports, and weekly watchlist digests. Read the docs. We also ship an MCP server for Claude, ChatGPT, and Cursor, and a Substack newsletter endpoint over a large, new dataset of finance newsletters.

Yes. Web Pro is $49.99 a month and removes the 20-view limit: unlimited transcript and feed views, keyword search across every executive, quotes, and AI chat in the web platform. It does not include the API, MCP server, or bulk exports; those come with the $499 Data API, which also includes the full web platform. See pricing.

Yes, and many teams do. Citation formats and a data availability statement are here.