Kscope · deep dive · 2026-10-18
RavenPack sells a daily score on the news. The score on a public story is a model. The score on a licensed wire is their feed.
The row stays fast. Company News, Controversy, Earnings Intelligence, Credit News, the C-level indicator, and the currency signals are models on text. The news archive and the structured events on it are already in house. The same scores on Dow Jones, the Journal, Barron's, MarketWatch, MT Newswires, FactSet, and the Financial Times stay a license. Bigdata adds broker research, estimates, and expert interviews on top of that license.
| Feature | RavenPack | Kscope | |
|---|---|---|---|
| Edge, the analytics record | Edge, read 2026-10-18, is structured analytics on published content. The page counts 12 million entities, 7.1 million companies, 4.5 million individuals, and 60,000 products and services. It claims 80-plus fields per entity, more than 20 sentiment indicators, and 7,000-plus event categories. Delivery is a dataset builder, a web API, or Snowflake. The same page says the archive is point-in-time, more than 20 years, and the service quotes 99.98 percent uptime. No seat price is printed. The door is a trial. | The news archive is in house: 30,176,943 documents as of the news note of 2026-09-19. A sentiment field on every entity in every story, and a 7,000-category event tag, are a model on that text. They are not shipped as a daily factor. | Partial |
| Factors, the daily score | The factor library, read the same day, packages the analytics into files a portfolio can import. The families on that page are C-level economic sentiment, forex sentiment, controversy news, company news, earnings intelligence, and credit news. Delivery is CSV, Snowflake, or a Python query. Each family page says request access. No dollar price is printed. | A daily score on the news, the filings, and the calls we already hold is a model. The structured events underneath it are already extracted. The smoothed factor file is the part still to build. | Partial |
| Bigdata, the agent store | Bigdata.com is theirs. The Financial Times release of 1 December 2025 says RavenPack launched it in 2024. The store page, read 2026-10-18, says it grounds an agent in 25-plus years of content. Public news is listed at 10 million new documents a month, with 5 rolling years. Filings are listed at 90,000-plus companies in 50-plus countries since 2010. Press releases cover 25,000 companies. A free account needs no card. Connectors named on the page include Claude, ChatGPT, Copilot, and Grok. | Search over the news archive and the US filings is live. A cited answer on that text is software on a corpus we hold. A cited answer that requires their premium catalog is the license in the next section. | Partial |
| Feature | RavenPack | Kscope | |
|---|---|---|---|
| The company-news factor | Company News, read 2026-10-18, scores media attention and sentiment for listed companies. The facts box says 6 times a day, history since 2001, and a universe of 100,000-plus listed companies. Document-level factors count mentions and average sentiment. Event-level factors do the same for a detected event. The chart on that page is labeled 2004-01-01 through 2023-08-01. The facts box does not say the series ends in 2023. | A mention count and an average sentiment, refreshed through the day, is a query plus a model on the archive. The history of that factor back to 2001 is a backfill across the documents we stored. | Partial |
| How much text | The same page says RavenPack analyzes over 300 million news and social items a month from 40,000 sources. Edge says the analytics also run in 13 languages, including English, Arabic, Chinese, Dutch, French, German, Italian, Japanese, Korean, Portuguese, Russian, and Spanish. | The stored archive on 2026-09-19 was 30,176,943 documents. 23,388,726 are RSS. GDELT is 139,355. A live newswire supplies the rest. The 2026 slice of that count was 8,005,734 documents through about 8.5 months. The corpus holds articles in about 40 languages. The ingest monitors about 100,000 RSS feeds. Their monthly figure counts news and social items. Ours is stored documents. | Partial |
| Events already pulled out of the news | Under the hood, Company News says a taxonomy sorts acquisitions and mergers, analyst ratings, assets, credit, credit ratings, dividends, earnings, equity actions, insider trading, labor issues, legal, marketing, partnerships, price targets, products and services, regulatory, revenues, sanctions, stock picks, and stock price. Edge says a sentiment score is still computed when no event is detected. | The structured layer on the same archive, counted 2026-09-19, is about 4.02 million events joined to the company graph: 1,248,515 appointments, 915,757 funding and merger events, 765,548 commercial relationships, 362,515 lawsuits, 320,855 dividends, 259,290 regulatory and government events, 53,412 personal windfalls, 37,492 data breaches, 37,198 bankruptcies, and 21,321 auditor changes. Analyst ratings, price targets, and stock picks are a separate opinion. | Partial |
| Feature | RavenPack | Kscope | |
|---|---|---|---|
| Four inputs, one score | Earnings Intelligence, read 2026-10-18, combines earnings news, transcripts, insider transactions, and, for US companies, earnings dates. The facts box says 6 times a day, at the open and the close in New York, Paris, and Tokyo, history since 2007, universe 100,000 listed companies. Insider volume between earnings announcements overweights or underweights the score. The page shows cumulative-return charts and does not print a return figure. | The four inputs are in house for US filers. The combined score, smoothed over 1, 7, and 30 days, with a 90-day or 180-day window, is a model on those inputs. | Partial |
| The call | The transcripts family on that page is sentiment on earnings calls for 20,000 global listed companies, 4 factor types, 24 values per snapshot. | The call file as of 2026-10-14 holds 264,816 call pointers, 189,500 joined to a company. The live file checked 2026-09-19 was 24,337 calls across 4,344 tickers, and 23,795 had a summary. A sentiment number arrives with the live transcript. The earnings-sentiment note of that day says the number has not been stored as a company factor. | Partial |
| The insider print, and the date | The insider family is one factor from executive stock transactions. The dates family, US only, is a daily read of the investor calendar, aimed at a change in the earnings schedule. | Form 4 and Form 5 are live: about 11.28 million classified transactions, 19,910 issuers, 209,003 people, insider note of 2026-09-19. Forward report dates are 3,690 forecasts as of 2026-10-05. A factor that tilts on net insider volume between two earnings dates is a model on those two files. | Partial |
| Feature | RavenPack | Kscope | |
|---|---|---|---|
| Controversy | Controversy News says it tracks controversial events for 7 million public and private companies. History since 2002. The score is positive and open-ended, and it decays as the story ages. The facts box says the file is daily. The construction section of the same page says the score is produced 7 times a day, at midnight UTC and at the open and the close in New York, Paris, and Tokyo. The factor library card on the index page says 6 factor types and 213 values per snapshot. | Lawsuits, breaches, bankruptcies, and regulatory events from the news are already structured. An ESG controversy score, with a decay and a pillar, is not in the graph. The ESG note of 2026-09-19 found no emissions, sustainability, or controversy score. | Partial |
| Credit | Credit News, read the same day, is two factors: mean credit sentiment, and credit event volume, across about 120 catalysts. History since 2001. The body says the universe is over 7 million public and private companies, and the factors are aggregated until 60 minutes before the credit close. The facts box on that page calls the entity a currency. The body calls it a company and a bond issuer. Their backtest description runs from September 2015 to February 2024. No return is printed in the text. | A credit-tagged event on a public issuer is a label on the news we hold, joined to the company graph. A score aimed at a bond spread, including private issuers, is a wider model. The letter a rating agency assigns stays with that agency. | Partial |
| The US economy, from the calls | C-level economic sentiment, read the same day, is one daily indicator for the US macro cycle, history since January 2003. It reads executive comments on consumer confidence, retail and vehicle sales, employment, payrolls, jobless claims, builders, permits, home sales, factory orders, and industrial production, then labels the phase expansion, recovery, slowdown, or contraction. The chart on the page runs through 2026-10-18. | The US call text is in the call file. A single smoothed macro series, and a phase label, is a model on that text. | Partial |
| Currencies | The factor index describes forex sentiment as daily trend-following and mean-reverting signals from macro news across the G7 currencies. Two factor types. Ten values a day. | Macro headlines are inside the news archive. A G7 trading signal on that text is a model. | Partial |
| Feature | RavenPack | Kscope | |
|---|---|---|---|
| The wires inside Edge | Edge names the premium sources: Dow Jones Newswires, Barron's, the Wall Street Journal, MarketWatch, MT Newswires, FactSet, LinkUp, and SCRIPTS Asia, plus a retail newswire, and over 40,000 web and social sources. The analytics on those wires are the record a public RSS item does not replace. | RSS, GDELT, the live newswire already in the archive, filings, and calls are the input the factor can use. A Dow Jones story that never hits a public feed is their copy. | Beside |
| The Financial Times | The 1 December 2025 release: FT Ventures invested, and the FT news feed plus an archive back to 2012 are licensed into RavenPack products, including Bigdata, for generative-AI use. The release names Armando Gonzalez as chief executive. It also names Bullhound Capital and Molten Ventures as investors. RavenPack describes itself there as more than two decades old. | The FT archive is a license. A model does not reproduce a story the FT has not put on an open feed. | Beside |
| The rest of the Bigdata catalog | The store page lists 200-plus premium sources since 2000, fundamentals it describes as 1 million-plus data points and 30-plus years including estimates, analyst research from 100-plus providers, 200-plus expert interviews a year through Knowledge Ridge, and 4,400-plus finance shows through Podchaser. Pricing read the same day: Starter is 0 dollars a month, with 10 dollars of credits once, and no card. Pro is printed at 40 dollars a month beside a 50 dollar regular price, with 50 dollars of credit a month. Max is printed at 160 dollars a month beside a 200 dollar regular price, with 200 dollars of credit a month. The yearly toggle says 20 percent off. Pro adds the premium catalog, estimates, expert interviews, and sentiment. | Filings, the call text, and the news archive ground an answer without that catalog. Estimates, broker research, and paid interviews stay a license. | Beside |
Ahead The feature is data we already hold, in a form an investor can use.
Match Both products do this.
Partial We hold the filing or the extract, and not their finished screen.
Gap They ship it. We do not.
Beside A licensed wire, a broker note, an expert interview, and a rating-agency letter stay theirs. A score computed from the news, the filings, and the calls we hold is a model.