Overview
The Earnings Call Transcripts listing delivers the verbatim text of more than 243,000 quarterly earnings calls from 6,650 U.S.-listed tickers, from October 2005 to today, ready to query in Snowflake. Each call is split into speaker paragraphs, so what the CEO said, or what analysts asked in the Q&A, is aWHERE clause away. Try it with the free Apple sample: all 85 Apple calls since 2005.
It’s built for equity analysts, quant researchers and data teams who want management commentary next to their own data in Snowflake, for text search, sentiment scoring or model training, without buying a feed and writing a parsing and loading pipeline.
Data is standardized into analytic‑ready views (one row per call with the full transcript, one row per paragraph, and a per-ticker summary), checked for new transcripts every day, and loaded only when the source changes, so you can query, join, model and build dashboards right away.
Get it on the Snowflake Marketplace
- Free Apple sample: every Apple (AAPL) call from 2005 to present, 85 transcripts
- 243,000+ transcripts from 6,650 tickers, as of November 2026
- History from October 11, 2005, with 18,000+ calls a year since 2022
- 15.6 million speaker paragraphs, about 59 per call, each with the speaker’s name and its position in the call
- Checked daily at 00:00 UTC; new transcripts usually reach the listing within 10 minutes
- Three views: full transcripts, paragraphs, and a per-ticker summary of counts and date ranges
- Works with Snowflake Cortex: the share also includes a semantic view and a Cortex agent for plain-English questions
What is earnings call transcript data?
Each quarter, most public companies hold a conference call after reporting results. Management reads prepared remarks on the quarter, often gives guidance, then takes questions from analysts. A transcript is the written record of that call, word for word. It gives you the reasoning behind the numbers in the company’s own words, plus the questions analysts chose to ask. What’s in each transcript, at a glance:- Prepared remarks: usually from investor relations, the CEO and the CFO
- Q&A: each analyst’s question and management’s answers, in order
- Speaker names on every paragraph, such as
Timothy CookorOperator - Fiscal year and quarter as the company reports them, plus the date of the call
Coverage
As of November 2026: 243,046 transcripts from 6,650 tickers. The earliest call is October 11, 2005 and the latest scheduled call is November 23, 2026.BF-A and BF-B, or MS-PA, and pairs like GOOG and GOOGL. These often carry the same transcript as the company’s main ticker, so filter them out, or deduplicate on the transcript, if you count companies.
Not included: companies without a U.S. listing, non-earnings events such as investor days, and speaker titles or firms (the speaker column has names only).
Schema
The listing has three views. Every transcript is identified bysymbol, fiscal_year and fiscal_quarter, and by earning_call_transcripts_key, an MD5 hash of those three, for example 05e83e1cc9bca60974fc2436a27c8e1b for Apple’s fiscal Q4 2025 call. Join earning_call_transcripts_flat_vw to earning_call_transcripts_vw on earning_call_transcripts_key, and either one to earning_call_transcripts_cnt_vw on symbol.
earning_call_transcripts_vw
One row per earnings call, with the full transcript as an array.
earning_call_transcripts_flat_vw
One row per paragraph of each call: (symbol, fiscal_year, fiscal_quarter, paragraph_number).
earning_call_transcripts_cnt_vw
One row per ticker, summarizing its calls. It covers every ticker in the listing, including during the free Apple sample, so you can see what the full listing has.
earning_call_transcripts_sv, and a Cortex agent, earning_call_transcripts_agent, built on these three views, so you can ask questions in plain English, such as “What did Tim Cook say in Apple’s Q4 2025 earnings call?”
Example queries
All example output below is real, pulled from the listing on November 20, 2026 at 04:23 UTC. The tariff search across all 15.6 million paragraphs took about 5 seconds on an X-Small warehouse; every other query finished in under 1 second. Each query runs as-is on the free Apple sample; the cross-company ones return only Apple’s calls until you buy the full listing.How many transcripts are in the free sample?
During the free sample, the transcript views return only Apple’s calls, so this count is the whole sample:What did Tim Cook say on Apple’s fiscal Q4 2025 call?
The first three paragraphs he spoke on the October 30, 2025 call, cut to 100 characters:How many transcripts are there, and how far back do they go?
The whole listing, then a few companies, from the summary view:How often did companies mention tariffs?
Calls with at least one paragraph mentioning tariffs, by quarter, across every company:Prepared remarks vs. Q&A
Because every paragraph has a speaker and a position, you can split a call in two. The first paragraph spoken byOperator opens the Q&A:
Data Ingestion
Data is ingested into Snowflake with best practices in mind.Change detection
Raw snapshot
Standardize
Merge only what changed
earning_call_transcripts_flat_vw.Data Quality
- Change detection: a new file is loaded only when its size or version differs from the last one loaded, so an unchanged file never reloads.
- Required fields: a transcript without a ticker, fiscal year, fiscal quarter, call date or text is never loaded.
- One row per call: calls are merged on ticker, fiscal year and fiscal quarter, so a call can’t appear twice.
- File lineage: every source file is logged with its size, version and load time, and kept in S3, so any load can be traced and replayed.
- Stops on failure: if any step fails, the run stops, later steps don’t run, and we’re alerted.
Disclosures
Fruit Stand Data LLC is not endorsed by or affiliated with any of the companies whose calls appear in the listing, or with the provider of the source transcripts. Use contact@fruitstand.dev for questions. Known gaps and quirks, passed through as the source publishes them:- Short or empty transcripts: 951 calls have 3 paragraphs or fewer, most from 2006 and 2007, often a placeholder with speaker
Unknown. 8 calls have no paragraphs at all, so they appear inearning_call_transcripts_vwbut not inearning_call_transcripts_flat_vw. About 48,000 paragraphs (0.3%) have emptycontent. - Shared transcripts: share-class and preferred tickers (see Coverage) often repeat the main ticker’s transcripts.
- Fiscal quarters follow each company’s calendar: Apple’s fiscal Q4 ends in September. A few companies have the same call under two fiscal years, because of how the source labeled them. Use
report_dateto compare companies in calendar time. - Speaker names only:
speakerholds a name, not a title or firm. Analysts can be found as the speakers right after anOperatorparagraph.
src_updated_at changes. Earlier versions aren’t kept.
F.A.Q.
Frequently asked questions from our clients.Is there a free trial?
Is there a free trial?
earning_call_transcripts_cnt_vw, shows every ticker, so you can check coverage for the companies you need. The full listing is $50/month flat, and nothing is charged unless you buy it.What are earnings call transcripts?
What are earnings call transcripts?
How often is the data refreshed?
How often is the data refreshed?
When does replication across cloud regions happen?
When does replication across cloud regions happen?
How far back do the transcripts go?
How far back do the transcripts go?
earning_call_transcripts_cnt_vw for any ticker’s first call.How is the data obtained?
How is the data obtained?
Why use this listing instead of a transcript API?
Why use this listing instead of a transcript API?
What does it cost to build this ourselves vs. buy the listing?
What does it cost to build this ourselves vs. buy the listing?
What isn't included?
What isn't included?
Can I request additional earnings call transcript data to be made available in Snowflake?
Can I request additional earnings call transcript data to be made available in Snowflake?