News Quantified vs RavenPack: Measured Reactions vs Sentiment Analytics
TL;DR
RavenPack and News Quantified answer different questions. RavenPack scores news as it happens: sentiment, relevance, and novelty across a very broad entity universe. News Quantified measures what markets actually did after classified news events: 26M+ equity news records over 20+ years, joined to measured price and volume reactions. If your research question is "how is this news worded and how prominent is it," RavenPack is built for it. If your question is "what has the market historically done after events like this," that is News Reaction Intelligence, and it is what NQ was built for. Many institutional workflows justify both.
The category difference
This is not two vendors selling the same thing at different prices. It is two different data objects:
- A sentiment record describes the news: this story about Company X is positive, highly relevant, and novel.
- A reaction record describes the market: after events of this class, here is the measured distribution of what stocks did on day 0 through day 5, across two decades.
Sentiment is an input measured at publication time. Reaction is an outcome measured after it. Correlating the first with returns is the researcher's job; the second already contains the returns.
Honest tradeoff table
| Dimension | RavenPack | News Quantified |
|---|---|---|
| Core object | Sentiment / relevance / novelty scores on news items | Classified events joined to measured market reactions |
| Historical depth | Multi-year archive, widely used in academic research | 20+ years, collected continuously since 2012 |
| Breadth of coverage | Very broad: large entity universe, multiple asset classes, macro and geopolitical themes | Deliberately narrow: US equities, corporate event classes, depth over breadth |
| Primary buyer question | "How is the news flow shifting, at scale, right now?" | "What is the historical base rate after this event class?" |
| Outcome labels | Not the product; researchers join returns themselves | The product; reactions are first-class fields |
| Event taxonomy | Broad taxonomies oriented to news content | Reaction-oriented taxonomy (scheduled/unscheduled flags, rumor chains) built for event studies |
| Academic validation | Extensive third-party literature | Peer-reviewed study of the dataset (Levi, Livnat, Zhang and Zhang, 2016) |
| AI-native access | API-centric | API plus MCP connector for direct AI-agent access |
| Where it is clearly better | Breadth, entity coverage, macro/multi-asset scope, brand ubiquity in the space | Reaction depth, outcome labels, point-in-time event discipline, IR-facing readouts |
Where RavenPack is genuinely stronger
Stated without hedging: if you need broad multi-asset news analytics, coverage of global entities and macro themes, or a sentiment factor with a long trail of third-party academic literature, RavenPack is the mature choice, and pretending otherwise would tell you something about the rest of this page. Breadth is their moat.
Where News Quantified is genuinely stronger
Depth on the reaction question. If the workflow is event studies, base rates, tail analysis by event class, backtests requiring outcome-labeled events with point-in-time discipline and delisted names included, or giving an AI agent queryable market memory, that is the dataset NQ has spent since 2012 building. Reaction depth is our moat, and it cannot be assembled retroactively; the label problem explains why.
Can you use both?
Yes, and sophisticated shops do exactly this pattern: broad sentiment coverage for flow monitoring, reaction data for event research and risk. The datasets join naturally on entity and time. This page exists to route your question to the right object, not to claim one object answers everything.
Frequently asked questions
Is News Quantified a RavenPack competitor?
Partially. Both operate in news-derived market data, but RavenPack's core object is the sentiment score and NQ's core object is the measured reaction. The overlap is smaller than the category labels suggest.
Which is better for backtesting event-driven strategies?
Backtests on event classes need outcome-labeled events with point-in-time timestamps and delisted coverage, which is NQ's design center. Sentiment-factor research at breadth is RavenPack's.
Which is better for a risk team?
For quantifying tail exposure by event class from measured historical distributions, reaction data is the direct tool. For monitoring live news flow at breadth, sentiment analytics are.
Does News Quantified score sentiment?
No. NQ classifies events and measures reactions. Tone scoring is deliberately out of scope; outcomes are the product.
Historical reaction data. Not investment advice. Past reactions do not determine future reactions.