Why use sentiment data for equity research?

Sentiment data measures what investors and the public are saying about a company, and how much they are saying it, in close to real time. Equity researchers use it to catch shifts in attention and opinion before they appear in prices, analyst revisions or quarterly fundamentals, and to explain moves that traditional data can’t.

Social media, forums and news are where retail and professional investors react first. Traditional equity data is either slow, like financial statements, or already priced in, like price and volume. Sentiment data sits between the two. It captures the conversation as it happens and turns it into numbers that can be screened, charted and tested.

What sentiment data is used for in equity research

• Early warning on narratives. A sudden rise in message volume (“buzz”) on a stock often comes before news coverage or a price move. Examples include a product problem spreading on Reddit or a short report circulating on X.
• Explaining price moves. When a stock moves without a filing or headline, the message stream usually shows what investors are reacting to.
• Earnings positioning. Discussion volume and tone in the days before an earnings release show how crowded expectations are.
• Risk monitoring. Negative sentiment shifts, unusual buzz and keyword events flag potential risks in a portfolio. Keyword events include bankruptcy, insider trading and pump-and-dump patterns.
• Systematic signals. Quantitative funds use aggregated buzz and sentiment time series as inputs to trading models, after backtesting them.
• Private company and consumer research. Private equity teams track social discussion of non-listed companies and consumer brands, where little structured data exists.
• Market surveillance. Exchanges and regulators use social data to detect manipulation, fraud and coordinated promotion. Research on 27 crypto rug pulls found coordinated bot activity around the scams, which may serve as an early warning sign (study).

Why the historical record matters

A sentiment signal is only useful for research if it can be tested against the past. That requires history collected at the time, not reconstructed later. Many original sources no longer offer historical views, and social platforms have restricted or monetised API access. As a result, sentiment history that was not captured when it happened is now mostly unrecoverable.

What independent research shows

• Sentiment and short-term equity returns (ESMA, 2024). The European Securities and Markets Authority studied STOXX 600 stocks from January 2019 to June 2023 using Stockpulse social media data. The sample covered about 300,000 observations across 580 stocks. Social media sentiment was significantly correlated with excess returns on the same day and the following day. The effect faded within days and was not significant at 5 or 10 days. For the most-discussed stocks, falling sentiment was associated with lower same-day excess returns. (ESMA TRV Risk Analysis, “Social media sentiment: Influence on EU equity prices”)
• Finfluencers drive crowd sentiment (Business & Information Systems Engineering, 2025). Researchers from the University of Cologne and Stockpulse analysed 80 million posts on X about stocks and cryptocurrencies from 2022, using Stockpulse data. Sentiment from influential accounts predicted crowd sentiment in the short term, but not the other way around. The effect was stronger for cryptocurrencies than for stocks, and stronger when the crowd was more divided. (Haase, Rath, Krauß, Schoder: “The Role of Finfluencers in Shaping Crowd Sentiment”)
• Bots in crypto scams (2023). An analysis of social media activity around 27 rug pulls found massive coordinated bot activity within and across the scams, much of it targeting established finance news outlets such as Bloomberg and Reuters. The authors conclude that bot deployment may be an early indicator of rug pulls and similar promote-hit-and-run scams. (On the Involvement of Bots in Promote-Hit-and-Run Scams: The Case of Rug Pulls)

Stockpulse sentiment data: key facts

What the data contains

• Buzz: a normalised measure of message volume for an instrument or topic.
• Sentiment: a contextual score from proprietary NLP models and large language models, at message, sentence and aggregated level.
• Messages: individual messages with source, language, entity tags, sentiment score, spam score and author score.
• Topics: weighted word clusters describing what is being discussed about an instrument.
• Key events: automatically detected corporate and market events.
• AI summaries: thousands of AI-generated instrument summaries per day.
• Alerts: real-time alerts in six categories: buzz anomaly, price-buzz combo, ad-hoc messages, pre-earnings buzz, delayed earnings and tracked users.

Who uses Stockpulse sentiment data

• Stock exchanges and regulators: Deutsche Börse (client since 2018), Nasdaq US and Nasdaq Nordic (since 2019, integrated into the SMARTS surveillance system), Bursa Malaysia, and more global and government institutions. They use it for trading surveillance.
• Financial data and research providers: Moody’s, Refinitiv and Interactive Brokers, which use AI-generated summaries and sentiment analyses.
• Researchers: Stockpulse data underlies peer-reviewed and regulatory research, including ESMA’s analysis of social media sentiment and EU equity prices and a study of finfluencers and crowd sentiment in Business & Information Systems Engineering.
• Investment firms: US and UK quantitative hedge funds use it for systematic alpha and risk monitoring, including strategies on China A-shares. A large US private equity firm uses it for deal sourcing and portfolio monitoring.

Frequently asked questions

Does sentiment data predict stock returns? It depends on the signal, the universe, the horizon and the construction. ESMA’s 2024 study of STOXX 600 stocks, based on Stockpulse data, found that social media sentiment was linked to excess returns on the same and following day, but the effect did not persist beyond a few days. Sentiment and buzz are inputs to be tested, not guarantees. Researchers should backtest any sentiment signal on point-in-time history before using it.

What is the difference between buzz and sentiment? Buzz measures how much an instrument is being discussed relative to its normal level. Sentiment measures whether that discussion is positive or negative. The two are often most informative in combination. For example, a buzz spike with sharply negative sentiment is a different situation from a buzz spike with neutral tone.

How far back does Stockpulse data go? Stockpulse has collected data continuously since 2011, giving more than 15 years of history.

Can I use sentiment data without writing code? Yes. The Stockpulse Dashboard provides charts, heat maps, watchlists and CSV exports. It also offers full-text search across every message collected over the last 15 years.

Contact: info@stockpulse.ai · www.stockpulse.ai