RavenPack Alternatives in 2026: What You Actually Get Below the Enterprise Tier

What RavenPack actually sells

RavenPack sells structured news analytics to institutions, not a sentiment number on its own. Its News Analytics page organizes the product under five headings in RavenPack's own terminology: sentiment analysis, relevance scoring, novelty tracking, temporal scoring and topic tagging. For every entity and event detected in a story, RavenPack attaches relevance scoring, novelty tracking plus impact analysis.

The company describes its approach as deliberately plural: it does not believe in "one size fits all" and has developed multiple sentiment techniques, some using millions of rule sets while others use machine learning. It dates its work in the field to 2003.

The commercial shape matters more than the feature list. RavenPack Analytics is a licensed dataset sold on a contract. The archive is a large part of what you are paying for. The catalog runs wider than news: job analytics, regulatory filings, transcripts, earnings dates, insider transactions and annotations. Prebuilt factors include Company News, Earnings Intelligence, Controversy News and Credit News. Macro indicators include a C-Level Economic Sentiment Indicator and Forex Sentiment.

The scale correction: CSS runs 0 to 100 and ESS pivots on 50, so neither runs -1 to +1

RavenPack's classic sentiment analytics run from 0 to 100 with 50 as the neutral point. The widely repeated claim that the Event Sentiment Score runs -1 to +1 is wrong for that analytic.

RavenPack's own Composite Sentiment Score research page on www.ravenpack.com, its 2011 research note, defines CSS as "a story-level sentiment analytic that represents news sentiment by combining various sentiment analysis techniques", states that "CSS can take values between 0 and 100" and describes the Event Sentiment Score as "positive if ESS>50 or negative if ESS<50".

ESS comes from a category list covering more than 330 market-moving events. The same page treats CSS as applicable once a company's relevance in a story passes 90.

The -1 to +1 scale is real. It belongs to Bigdata.com, whose API reference documents a daily_sentiment field that "ranges from -1 (negative sentiment) to 1 (positive sentiment)". The reason to get this right is a filter that breaks silently: a rule written as score > 0 means bullish works on the Bigdata scale, while on the CSS scale it matches every story ever scored, because the CSS scale never goes below 0.

Coverage: what ravenpack.com claims today versus what only third parties claim

Four coverage numbers are verifiable on ravenpack.com itself:

• Recognition of "over 12 Million named entities"
• A proprietary taxonomy covering "more than 7,000 different event topics"
• "more than 40,000 sources of news and social media in 13 languages"
• The statement that "80+ fields describe every entity detected including over 20 sentiment indicators"

Bigdata.com carries its own set: 170+ premium providers, 100+ countries, 25+ years of history and 1B+ documents, with the news dataset described as content since 2000, regulatory filings covering over 50 countries since 2010 and web data held as five rolling years.

Note what a language count does not tell you. RavenPack advertises 13 languages, LSEG MarketPsych 12 and Permutable.ai 80+. None of the three says whether non-English text is scored natively or machine-translated first, which are two different measurements.

A second set of figures circulates on aggregator pages and academic data catalogs without appearing on ravenpack.com today: 45,000 companies, 143 countries, 350,000+ entities, 25,000+ sources and "20+ years with millisecond precision". That last one deserves the same scrutiny as the rest. Millisecond precision back to 2003 would require the original wire feeds to have carried millisecond stamps. RavenPack makes no such claim on its own site today.

The history start date is the sharpest case. WRDS at Wharton lists the Dow Jones, Full and Web editions of RavenPack Analytics as running 2000-01-01 to 2026-06-30, while RavenPack's own page promises "full historical and real-time updates from premium sources" with no year attached. If a start year decides your purchase, get it from the vendor in writing instead of from a listicle.

Their own pages also disagree with each other in two places. Cite one page and never average. The News Analytics page says 12 million named entities while the homepage says 7M+ companies, different units that are easy to misread. One catalog page carries both 170+ premium providers and "200+ Premium licensed sources".

On the customer side, RavenPack states that "more than 70% of the best performing quantitative hedge funds and asset managers use RavenPack News Analytics for alpha generation and risk management". That is the company's own unaudited claim rather than an independent finding, so treat it as marketing.

What the enterprise contract buys that no website lists

Coverage counts are not the deliverable. Three line items decide what a RavenPack-class contract is worth, none of which appears on a public page.

The entity master. Bigdata.com's sentiment endpoint still returns RavenPack entity IDs. An identifier layer sits underneath everything else. Read that as a deliverable, not as a coverage stat.

What has to be in the contract is the mapping history: historical ticker, ISIN and CUSIP mappings, listing histories and entity lifecycle events. RavenPack does not document any of that publicly. Put it in the RFP. It is also why leaving hurts: your joins end up keyed to a vendor's identifiers.

Versioning and restatement. Ask whether analytics carry a version and whether history is reprocessed when a model is upgraded. No vendor on this page publishes a restatement policy. The answer only ever arrives in a sales call. Get it in writing. A silently restated archive turns a reproducible backtest into an unreproducible one, which is an operational risk in any production signal.

Delivery, support and incidents. RavenPack describes bulk history and feeds for systematic use. Bloomberg sells through its enterprise data catalog. Beyond that, none of the three enterprise vendors publishes its channel list, its support SLA or its data-quality incident process.

Integration channel decides these deals more often than price does. Name yours in the RFP: file drop, streaming feed, warehouse share or hosted research environment are four different engineering bills. Ask how fast you are told when a source disappears from the feed.

Bigdata.com is RavenPack, but it is not a replacement for RavenPack Analytics

Bigdata.com is a brand RavenPack owns, launched in 2024 as an agentic AI platform, with a footer that credits "Bigdata.com by RavenPack". The two sites position differently. ravenpack.com leads with technology and insights for data-driven companies. bigdata.com calls itself "the AI grounding layer for finance" that grounds any AI agent in premium business and financial content "so it stops hallucinating".

Three differences decide whether it substitutes for what you were about to buy.

• Buying model: RavenPack Analytics is sales-gated with no public price, while Bigdata.com is self-serve and metered per token, which lets a developer start without talking to anyone.
• Deliverable: RavenPack sells a structured analytics dataset built for systematic work (CSS and ESS, relevance, novelty, 0 to 100 scales, bulk and feed delivery), while Bigdata.com sells retrieval and grounding for AI agents, "delivered as ranked, cited excerpts your agents can trust, never raw, unverifiable dumps".
• Both are still sold: RavenPack's navigation, product pages and a December 2025 press release still market the analytics datasets alongside Bigdata.com.

A quant buying for backtests is still buying RavenPack Analytics on an enterprise contract.

Why nobody can tell you what RavenPack costs

RavenPack publishes no pricing anywhere. The URL www.ravenpack.com/pricing returned HTTP 404 on 26 July 2026, checked on the canonical www host and not on the apex domain, where deep paths serve a generic edge error that proves nothing. What comes back on www is RavenPack's own rendered 404 page, which is the site itself saying the page does not exist.

There is no tier list, no rate card, no self-serve path. Every call to action across the site is "Request Trial" or "Get Started", each landing on the same lead-capture form, which instructs you to "please use your business email". Any dollar figure you find for RavenPack on a comparison site is therefore unsourced. We are not going to invent one here.

Quote-only procurement is a different purchase from an API subscription. Four things change.

• You see a number after a sales cycle, not before. Evaluation runs through a gated trial rather than a free tier, so the cheapest way to disqualify a vendor is to read its documentation before you fill in the form.
• The price is a function of your use case. Seats, delivery method, asset classes and whether the output leaves your firm all move it. There is no list price to hold up against a competitor's.
• Contract length is not published. Do not assume an enterprise feed is month-to-month. Ask for the initial term, renewal window and termination rights before evaluating the headline fee.
• Redistribution is negotiated separately. Whether you may show scores to end users, store them or resell a signal derived from them is a contract term. Budget the legal review, not just the fee.

Bigdata.com is the one part of the RavenPack estate with a pricing page, yet it still does not publish a rate. The page says the company charges "per token consumed, from fractions of a cent" with "no seats, no subscriptions to start". It offers "$50 in free credits" with "no credit card required".

Enterprise buyers get the familiar line, volume pricing and data licensing "tailored to your firm" behind a "Talk to us" button. The dollar amounts on that page, $0.57 to $1.82 per workflow run and $1.70 for an NVIDIA initiation report, are outputs of an interactive cost estimator rather than a published rate card.

Counted from that table: six vendors publish a rate card you can read in a browser, one publishes a free allowance with no unit rate and six publish nothing at all. That ratio, not the feature matrix, is what decides who can actually buy which product.

Three tiers: which one are you on?

News sentiment vendors are not interchangeable logos. What separates them is the data itself: whether a licensed multi-decade archive sits behind the score, whether that archive is point-in-time and what the license lets you do with the output. Procurement friction follows from those three things.

RavenPack, Bloomberg and LSEG MarketPsych keep prices behind a sales process because the content is licensed and an archive of that shape gets delivered on negotiated terms. Bigdata.com, Context Analytics, Permutable.ai and Benzinga each sell one narrower signal. Marketaux, Alpha Vantage, EODHD, Finnhub, StockGeist and Adanos take a card today. Work out which rung that puts you on before you open a vendor page.

Rung one: enterprise, quote-only, deep archive

The enterprise rung is RavenPack, Bloomberg and LSEG MarketPsych. All three sell licensed data on negotiated contracts, none of them publishes a price and none of them lets you sign up with a card.

Bloomberg sells news sentiment inside its enterprise data catalog. There is no standalone product to buy. The Event-Driven Feeds and Real-Time News Feeds pages describe analytics that "include sentiment analysis, novelty scores, readership heat, social velocity and more".

A new set of Real-Time News Feeds launched on 3 March 2026 adds "story-level Company Sentiment scores and Market Moving News indicators", which Bloomberg says are "powered by proprietary deep learning models trained on Bloomberg News". A separate ESG-specific product, ES News Sentiment, uses "a machine learning model that has been trained by Bloomberg analysts".

One correction worth making: MARS is a Bloomberg risk system, covering valuations, market and credit risk, XVA and collateral among others. Its product page mentions no news sentiment at all, so do not credit MARS with a sentiment feed. Bloomberg's own pages disagree on coverage counts, 151 against 146 bureaus and 10,000+ against 5,000+ stories a day. Quote one page. History depth is not published.

Bloomberg publishes no price either. The third-party figures in circulation for a Terminal seat contradict each other. The obstacle is not usually price. Bloomberg sentiment arrives inside an enterprise data relationship where what you may redistribute is negotiated instead of published, which is why Bloomberg is seldom the answer to "replace RavenPack".

LSEG MarketPsych is the most forthcoming of the three about what you get, if not about what it costs. Its 2026 brochure describes a natural language processing engine that turns global news and social media into structured indicators, running on a history from 1998 across 4,000+ curated sources, with 50+ sentiment scores updated every minute, hour and day. The separate ESG factsheet covers 252 countries and 12 languages.

The current brochure does not publish price, minimum term or redistribution rights. Those are contract questions, not facts to infer from the product sheet: ask for the initial term, renewal mechanics, storage rights and permitted downstream use.

Who the enterprise rung fits: a fund or a bank that needs a multi-decade archive for a long backtest, licensed redistribution, a vendor its compliance function has already onboarded and a budget line that survives a contract commitment. If that describes you, treat the rest of this page as context, not as a shortlist.

Rung two: mid-market and specialist

The mid-market rung is Bigdata.com, Context Analytics, Permutable.ai and Benzinga. Each sells one specific signal instead of a general-purpose analytics archive. Pricing here is semi-published at best: only Bigdata.com shows any number, the other three quote. Two of the four, Permutable.ai and Benzinga, do not belong on a RavenPack shortlist at all, for reasons worth reading before you book a vendor call.

Bigdata.com is the only self-serve door into RavenPack's content: $50 in free credits, no card, per-token metering. It is counted here for completeness, not as an independent vendor, because it is RavenPack's own second brand. Its entity-sentiment endpoint returns daily sentiment, sentiment pressure and abnormal media attention on a -1 to +1 scale, weighted by relevance and novelty.

Those three sentiment measures are a narrower surface than the 80+ fields RavenPack advertises on its own analytics, so treat Bigdata.com as a different product, not as a cheap RavenPack.

Context Analytics scores social posts, not news. Formerly Social Market Analytics, the company renamed in November 2022. BridgeWise announced its acquisition on 17 February 2026. The site now describes itself as a BridgeWise company. The deal value has not been disclosed that we could find.

Its S-Factor and S-Score products read X and StockTwits, with S-Score values from -4.25 to +4.25, one-minute updates and 3,000+ securities tracked. The clearest evidence of history depth on the site is a backtest chart captioned "Price > $5 Universe: 2011-12-01 to 2025-08-18", which points to a December 2011 start. The company is still shipping, with a podcast sentiment product added in June 2026. Note the live domain is contextanalytics-ai.com, with the hyphen.

Permutable.ai does not replace RavenPack for single-name research and does not claim to. This is a London vendor selling macro, commodity, currency and geopolitical sentiment indices, not equity-ticker sentiment. A shortlist built around company news is the wrong shortlist for it.

Its seven-stage Intelligence Engine is self-described as point-in-time by construction, with directional and semantic sentiment plus source-level traceability, 1m+ narratives processed daily, 250k+ sources, 80+ languages, 70+ assets and 11+ years of point-in-time history. Treat "point-in-time by construction" as the vendor's claim until check 5 below settles it.

Its own pages disagree on country coverage, 95 on the homepage against 90+ on the API page. Pricing is quote-only: the /pricing/ URL resolves to the homepage. A June 2026 release confirms the company is active.

Benzinga should come off a sentiment shortlist until it answers one question. The NewsQuantified API description claims sentiment scores, relevance metrics and sentiment polarity, but the endpoint's own documented 200 OK sample response carries 88 fields and not one of them is a sentiment field. Those fields are news-reaction price and volume analytics.

The closest editorial product Benzinga documents is WIIMs, "the one sentence reason an asset is trading higher or lower on any given day". What Benzinga does publish is scale: a historical news archive of 84 terabytes, Wilshire 5000 coverage, datasets since 2009 and latency of "less than 0.1 seconds". Pricing is quote-only. A free API key exists, with limits we could not verify.

Who the mid-market rung fits: teams that want one specific signal instead of a general archive, with enough process to survive a quote cycle. Read the deliverable here. The category label tells you nothing.

Rung three: developer self-serve

The self-serve rung is Marketaux, Alpha Vantage, EODHD, Finnhub, StockGeist and Adanos. Every one of them publishes a price and takes a card, with no quote cycle in between. Every one of them also publishes either a short history window or no history depth at all. None of them advertises a point-in-time research archive.

Marketaux returns per-entity sentiment "between -1 and +1", where 0 is neutral, defined as the average sentiment of all highlighted text found for the identified entity. Coverage is 200,000+ entities, 5,000+ news sources, 80+ global markets and 30+ languages. History depth is not published.

One limit to plan around: full article text is not included, because Marketaux provides "only a short snippet of articles along with their links". Free is 100 requests a day with three articles each, then $29.00 a month for Basic, with annual billing described as two months free.

Alpha Vantage returns an overall sentiment score plus a ticker sentiment array carrying a relevance score per ticker, with label bands defined inline in every response, from Bearish at -0.35 or below through Neutral between -0.15 and 0.15 up to Bullish at 0.35 or above. It tags 15 topics. Source count and history depth are not published.

The useful detail for anyone testing on zero budget: the news sentiment endpoint is not premium-gated, so it works on the free tier of 25 requests a day. Paid plans start at $49.99 a month for 75 requests a minute.

EODHD gives you a per-article breakdown with polarity plus negative, neutral and positive components. A separate daily aggregate endpoint returns a normalized value from -1 to +1.

Prices are in euros: EUR 19.99 a month for EOD All World, the same EUR 19.99 for the standalone news API, up to EUR 99.99 for the all-in-one plan. The free tier reads as 20 API calls a day, but each news request consumes 5 calls plus 5 per ticker, so in practice it is about two single-ticker news requests a day.

Finnhub is the vendor most often mis-described in this category. Read its docs before you budget. Its news sentiment endpoint is marked "Premium Access Required" and is documented as available only for US companies. Its social sentiment endpoint is marked "Premium required." Neither is in the free tier, which covers 60 API calls a minute of US market data for personal use.

News Sentiment sits on the Fundamentals tiers from $50 a month per market. Social Sentiment and Filings Sentiment sit only on the All-In-One plan at $3500 a month billed annually. Anyone who tells you Finnhub gives away sentiment for free is reading an old page.

StockGeist is publicly degraded at the time of writing, so treat it as a maybe rather than as a live option. A sitewide banner on every page reads "StockGeist.ai dashboard is under maintenance." Its documentation host fails TLS. Its API host returned HTTP 200 with a zero-byte body on every path we probed. The company has not been acquired or wound down. It is owned by Neurotechnology. How long the banner has been up is not published.

On the product itself, its two pages disagree on coverage, 2200+ publicly traded companies on the homepage against more than 2000 US stocks on the API page. History is short and tier-gated: 24 hours at five-minute granularity free, 7 days hourly on Starter, 30 days daily on Pro. Billing is credit packs, not subscriptions.

Adanos is ours, so read the disclosure below before you weigh it. It is a self-serve API covering five sentiment sources on one key: Reddit stocks, Reddit crypto, X, stock news and Polymarket. The news dataset refreshes every 10 minutes across 35,000+ tracked tickers. Scoring is a finance-tuned RoBERTa and VADER ensemble weighted 60/40 with 155 finance-specific lexicon terms, scored -1 to +1 per mention.

Free is 250 requests a month. Hobby is $29 a month. Professional at $299 a month carries commercial use, raw mention data and up to 365 days of history. That 365-day ceiling is one of several reasons Adanos sits on this rung and not the first one.

Who the self-serve rung fits: a product team shipping a feature, a solo quant testing an idea on a recent window, an agent or a bot that needs a live score. If your research has to cross 2008 or even 2015, none of these vendors documents an archive that reaches it.

Thirteen news sentiment APIs compared by tier, pricing model and history depth

Six of the thirteen vendors below publish a price you can read without contacting sales. The split runs almost exactly along the tier boundary. Rows are sorted by tier, enterprise first, then mid-market, then developer self-serve, in the same order as the profiles above.

Rows are not interchangeable across tiers. A 365-day window and an archive starting in 1998 share a column here because the column has to hold both, not because they are the same quantity. The sentiment-method column is where most comparison pages start. It is also the least decisive one: every vendor here scores text, while only some will sell to you on terms you can accept.

Four continuity events in this set and the three contract terms they imply

Vendor mortality in news sentiment is high enough to be a procurement input. Four of the names on this page changed hands, changed shape or went dark inside eighteen months.

• Wand AI acquired Accern in February 2025, a deal Forbes dated 7 February 2025 and reported as eight figures with the amount undisclosed.
• BridgeWise acquired Context Analytics in February 2026.
• StockGeist has been running a sitewide maintenance banner with a documentation host that fails TLS.
• RavenPack took an investment from FT Ventures in December 2025 and now markets a second brand beside its datasets.

None of this is scandal, just the base rate in this category, which makes it a contract question. Three terms follow.

• Insist on a perpetual license to data received during the term. A change of owner then cannot retroactively strip out the history your research runs on.
• Insist on bulk export on termination, in a documented format, agreed before you sign.
• Do not sole-source a production signal without a documented fallback. Treat dependence on any one feed as operational risk instead of reading exclusivity as a mark of quality.

The same base rate explains why a 2026 comparison page that still lists Accern as a live option has not been rechecked.

Both accern.com and www.accern.com fail the TLS handshake with no certificate served at all, reproduced across three separate clients. The Wayback Machine shows captures through 7 February 2026 and none after 15 February 2026. No vendor notice explains it, so we are not calling it a shutdown. We are calling the page that lists it unverified, which is a reasonable prior for the rest of that page.

A fourth route: WRDS, if you have a university affiliation

For academics the procurement answer is neither enterprise nor self-serve. It is WRDS at Wharton, which lists the Dow Jones, Full and Web editions of RavenPack Analytics as running 2000-01-01 to 2026-06-30. A large share of published RavenPack research reaches the data that way, so ask your library before you ask a sales team. Adanos runs a non-profit and education access program for the same population.

What to check before you sign anything

Choosing a news sentiment provider comes down to eight checks that apply at every tier, from a $29 API to a quoted enterprise contract. Run them in this order. The first one measures something. The other seven decide who is worth measuring.

1. Measure a sample against outcomes you already have

Every other check on this page is a question you ask a vendor, the kind of thing vendors are practiced at answering. This one is the only measurement on the list. Run it early, while the sales process is still cheap.

• Request a sample file covering a period for which you already hold labeled outcomes.
• Join it to your own universe on your own identifiers.
• Measure hit rate, how fast the signal decays and how much turnover it would have forced.
• Score it against a naive baseline: raw article counts on the same names over the same window.

A scored sentiment feed that cannot beat counting articles is an expensive counter. Everything below is a filter for deciding who you bother running that test with.

2. Entity tagging quality

A sentiment score is only as good as the link between a story and a ticker. Ask how the vendor resolves ambiguous company names, whether tagging is versioned and what happens to a story that mentions six companies. RavenPack sells this as a first-class surface, entity detection with relevance attached. Some cheaper feeds return a single article-level score with no entity resolution at all, which is a weaker signal wearing the same name.

3. Does each score carry relevance and novelty

Ask whether each score carries a relevance value and a novelty value, because both change the number you act on. Without relevance, a passing mention weighs the same as the subject of the story. Without novelty, one wire story syndicated to forty outlets reads as forty events, which is exactly the artifact RavenPack's novelty tracking exists to remove.

RavenPack publishes a Relevance score and treats CSS as applicable above relevance 90. Alpha Vantage returns a relevance score per ticker. Marketaux averages sentiment over the highlighted text for an identified entity. Our own BuzzScore counts mention volume as one of its five components with no novelty measure published beside it, which is a limitation to price in when one story is everywhere at once.

4. History depth against your backtest window

Write down the window your research needs before you open a single vendor page, because this is the check that sorts the three tiers. LSEG MarketPsych publishes 1998 to the present. Context Analytics shows backtest evidence from December 2011. Permutable.ai claims 11+ years point-in-time. Adanos caps at 365 days. No amount of feature parity closes the gap between a twenty-year window and a twelve-month one.

5. Point-in-time correctness, tested and not asked

"Point-in-time" stays a marketing phrase until somebody diffs two files. Permutable.ai advertises its engine as point-in-time by construction, which is the claim. The test is a diff: ask for the historical file exactly as it was delivered on a past date, diff it against today's archive for the same window and count the changed rows. That is a five-line email and it settles the question permanently. A vendor that cannot produce the old delivery has already answered.

Two questions travel with it. Timestamps come in three clocks: when the event happened, when the wire published it and when the vendor made the record available to you. Only the third is an honest backtest clock, so ask which one the field holds.

Look-ahead can live in the labels as well as in the archive. Ask for the model's training cutoff and how labels were created: outcome-derived labels or evaluation examples that overlap the backtest window can contaminate results even when the archive is genuinely as-published. General pretraining on contemporaneous text is not, by itself, proof of leakage.

6. Survivorship and entity mapping over time

Two failure modes quietly corrupt a joined backtest. Neither appears in a feature matrix. The first is survivorship inside the archive. Source lists change over an archive's life. A 2005 slice of a 2026 archive reflects the sources of 2005, which makes mention counts incomparable across time until you normalize for it. Ask whether delisted and bankrupt names stay in the entity file and whether their stories stay in the archive.

The second is entity mapping over time. FB became META. A ticker freed by a delisting gets reassigned to an unrelated company. A merger collapses two entity IDs into one. A spin-off splits one into two. Ask whether the vendor keeps a dated mapping table you can join as of a given date, because applying today's mapping to a 2012 story is the most common way a news sentiment integration produces confidently wrong numbers.

7. Redistribution and derived-data rights

Showing scores to your end users, storing them and reselling a signal derived from them are three separate permissions. Each is a contract term, never a feature. On the enterprise rung it is negotiated. On the self-serve rung it is usually a tier, where the cheaper plans grant personal use only. Read the plan text, not the marketing page. Settle this before a product roadmap depends on it.

8. Delivery method against your stack

Bulk history files, a streaming feed, a REST API and a hosted research platform are four different integrations with four different engineering costs. RavenPack sells feeds and bulk history for systematic use. Bigdata.com sells a research assistant plus an API. The self-serve vendors sell REST. Picking the wrong delivery shape costs more engineering time than the license fee saves.

Where Adanos fits and where it does not

One of the thirteen products above is ours, which is a conflict of interest best answered with specifics. Adanos sits on the developer self-serve rung. For a buyer who needs what RavenPack actually sells, it is not an alternative at all. The two columns below mark exactly where that line falls.

If you need a multi-decade licensed archive with negotiated redistribution, staying with RavenPack is very likely the correct decision. The point of this page is that you can tell which case you are in from published facts, before a sales call instead of after one.

FAQ

Sources

Pricing and feature notes captured July 2026 from each vendor. Tools change their plans, so verify current terms before you commit. Entries are marked by origin. A claim sourced to ravenpack.com or bigdata.com is the vendor's own. A claim marked third party is not verifiable on the vendor's site today. Where an entry carries no link, our capture holds the page and the quoted text without a stable URL, so it is cited by name.

1. RavenPack, Composite Sentiment Score and Event Sentiment Score, a 2011 research note: the 0 to 100 scale, the 50 neutral point, the relevance-90 rule (vendor page)
2. RavenPack, News Analytics: entity detection, relevance, novelty, 12 million entities, 7,000 event topics, 40,000 sources in 13 languages (vendor page)
3. RavenPack homepage: the 7M+ companies count and the "80+ fields describe every entity detected including over 20 sentiment indicators" line (vendor page)
4. RavenPack pricing URL, which returned HTTP 404 on 26 July 2026 on the canonical www host (vendor page)
5. Bigdata.com by RavenPack: positioning, headline coverage stats, the pricing page with $50 free credits, per-token metering and the cost-estimator figures (vendor pages)
6. Bigdata API reference, entity sentiment: the -1 to +1 daily sentiment field, sentiment pressure, abnormal media attention, quadratic relevance and novelty weighting (vendor docs)
7. Context Analytics: S-Score range, one-minute updates, securities tracked, the BridgeWise ownership line, the backtest chart caption (vendor pages)
8. Bloomberg enterprise data catalog, Event-Driven Feeds: sentiment analysis, novelty scores, readership heat and social velocity (vendor page)
9. LSEG MarketPsych 2026 brochure: history from 1998, 4,000+ sources, 50+ sentiment scores and 60-second updates (vendor document)
10. Permutable.ai Market Intelligence API: source counts, point-in-time claim and 11+ years of selected history (vendor page)
11. Benzinga NewsQuantified API documentation: the endpoint description and documented sample response (vendor docs)
12. Marketaux pricing: coverage counts, free limits and plan prices (vendor page)
13. Alpha Vantage NEWS_SENTIMENT documentation: score fields, label bands and topic list (vendor docs)
14. EODHD financial news API documentation: article and aggregate sentiment fields (vendor docs)
15. Finnhub sentiment documentation and pricing: premium markings and free-tier scope (vendor pages)
16. StockGeist API pricing: maintenance state, coverage counts, tier history windows and credit packs (vendor page, degraded at capture)
17. WRDS at Wharton, RavenPack dataset listing: the 2000-01-01 to 2026-06-30 range for the Dow Jones, Full and Web editions (third party)
18. Forbes, 7 February 2025, on Wand AI's acquisition of Accern (third party)
19. BridgeWise resources: acquisition of Context Analytics announced 17 February 2026 (buyer page)
20. RavenPack, 1 December 2025: FT Ventures investment and content partnership (vendor page)
21. Internet Archive Wayback Machine captures of accern.com (third party, no URL captured)
22. Adanos, financial news sentiment API roundup with the scoring methodology behind our own numbers
23. Adanos BuzzScore whitepaper: the ensemble weights, lexicon counts and per-mention scale cited for Adanos

Related