RavenPack News Analytics
Entity-level sentiment and event tags over 40,000 news sources, licensed by contract.
by RavenPack
Last updated
What it is
The dataset RavenPack has sold to quantitative funds since long before anyone called this category AI. It reads news and social content, detects the entities and events in it, and emits a row per detection rather than an article — which is the whole point, because a row is something a factor model can join on.
News Analytics is one of seven datasets inside RavenPack Edge, beside Job Analytics, Regulatory Filings, Transcripts, Earnings Dates, Insider Transactions and Annotations. Edge's separately sold Factors — Company News, Earnings Intelligence, Controversy News, Credit News, plus macro indicators on C-level sentiment and forex — are derived signal products built on top, and are not this card.
Pricing
Nothing is published. There is no pricing page on the site, and every call to action on the product page and the Edge page is the same "Request a trial" form, which refuses a personal email address. A trial exists; its length is not stated. Because the datasets are licensed individually and delivered together, the figure you get back depends on which ones you take and how much history you want.
Data & coverage
More than 40,000 news and social sources in 13 languages — English, Arabic, Chinese in both scripts, Dutch, French, German, Italian, Japanese, Korean, Portuguese, Russian and Spanish — including gated premium content from Dow Jones Newswires, the Wall Street Journal, Barron's, MarketWatch, MT Newswires, FactSet, LinkUp, Benzinga and SCRIPTS Asia.
The entity map is the asset. RavenPack claims 12 million named entities: 7.1 million companies public and private, 4.5 million individuals, and over 60,000 products and services, with a reference map of security symbols behind it. The event taxonomy runs to more than 7,000 categories with sub-taxonomies for macro, ESG and earnings, and a sentiment score is computed for a detected entity even when no event fires. Every detection carries 80-plus fields and more than 20 sentiment indicators — overall, sustainability, credit, interest-rate and risk impact — plus relevance, novelty and temporal scoring. The archive is claimed at 20-plus years, fully normalised and point-in-time, on a service the vendor puts at 99.98% uptime.
The vendor's solution pages market the same analytics for equities, fixed income, FX, commodities and crypto. What the dataset itself contains is tagged text about entities, so the asset-class framing is yours to build, not a field you are handed.
Integrations
Three deliveries, chosen at order time: a dataset builder for files, a web API, and Snowflake.
The Python client ravenpackapi is the vendor's own, and quiet — 1.1.6 in July 2024, last commit
October 2024. There is no free key, no sandbox and no public endpoint to poke at; the API
documentation sits behind the client portal.
Limitations
- No public price, no public terms and no public docs. You cannot size this, read the redistribution clause, or check an endpoint's shape before talking to sales. That is normal for the tier it sells into and a real cost if you are one person with a hypothesis.
- You are buying analytics, not a corpus. The premium half of the source list is gated content whose publishers licence it to RavenPack, not to you. Assume showing that text to your own users is a separate negotiation with each publisher, and get the display rights written down.
- The archive's start date is not published, only "20+ years" — and for point-in-time backtesting the first date and the revision policy are the facts that decide whether the data is usable.
- Latency is an adjective. The Edge page says "low latency API" and gives no number.
- The Python client has not shipped a release since July 2024.
- Thirteen languages is broad, but it is a list, not global coverage — and the sentiment models are trained on financial language, so a score on a non-financial story is a score on something the model was not built for.
Alternatives
Bigdata.com is the same company's self-serve product and the only way to touch RavenPack's index with a credit card. If per-entity sentiment over news is the thing you want and a contract is not, Marketaux sells it from $29 a month and Stock News API from a published price list. Benzinga News API is the other licensed newswire in this catalogue with no published price — same shape of negotiation, narrower to US equities.
Compare it against the rest of the market data APIs.
Specs
- Interfaces
- API, Python
- Export
- CSV, JSON
- Asset classes
- Stocks
- Markets
- Global
- Platforms
- Web, Library
- AI features
- None
- Capabilities
- News
- Pricing verified
- Capabilities verified
- Coverage verified
Also from RavenPack
Also worth comparing
- Benzinga News API — Licensed US stock newswire over REST, TCP, WebSocket or webhook, priced by contract.
- finlight — Financial news API resolving companies to tickers, ISINs and FIGIs, plus a stream.
- Marketaux — Global financial news over REST, tagged to entities and scored for sentiment per entity.
- NewsAPI.ai — Global news API with Wikipedia-grounded entity tagging and event clustering.
- NewsFilter.io — Ticker-tagged news API with a websocket push stream and a stalled SEC filings source.
- Stock News API — US stock news with a keyword sentiment label — headline and link, never the article text.
FAQ
Did RavenPack rename itself Bigdata.com?
No. RavenPack launched Bigdata.com in October 2024 as a second product and still sells Edge under its own name — the site's product menu lists the seven Edge datasets, then links Bigdata.com out separately, and the 2026 copyright line on both sites says RavenPack. They are two products with different audiences and different pricing models. See [Bigdata.com](/tools/bigdata-com) for the self-serve one.
Does News Analytics give me the articles themselves?
Not as the product. What is sold is structured analytics over the text — an entity or event detection with 80-plus fields describing it, including sentiment, relevance, novelty and temporal scores. Whether you may also see the underlying premium text from Dow Jones, the Wall Street Journal or Barron's is a licensing question with those publishers behind it, and nothing public answers it.
How far back does the archive go?
The vendor claims "20+ years of high-quality, fully normalized archive with billions of documents" and no start date anywhere public. That matters more here than on most cards, because a point-in-time archive is most of what a quant buyer is paying for — get the exact first date, per source, in writing.
Is there a Python client?
Yes, and it is the vendor's own — ravenpackapi on PyPI, published from dataservices@ravenpack.com, with source at github.com/RavenPack/python-api. It is quiet — release 1.1.6 shipped on 10 July 2024 and the last commit to master landed on 25 October 2024.