A research platform that answers a simple question: what's popular and why? We collect the public conversation about more than 14,000 stocks, ETFs, and cryptocurrencies across 15+ years from social networks and forums. Once we filter out the spam, we put over 100 proprietary models to work analyzing and visualizing the conversation and how it's changed over time.
We do not provide investment advice, but we believe the best use of our data is as one input of many in a thoughtful investment thesis.
The key to success is being where other investors will be, several months before they get there. So ask yourself whether the story about your stock is one that is likely to catch the fancy of the crowd. Can the story generate contagious dreams? Is it a story on which investors can build castles in the air—but castles in the air that really rest on a firm foundation?
Public posts from across the internet. This includes popular social networks as well as smaller forums and communities. We are always adding new sources of data and when we do, we update all of our metrics to reflect the most up-to-date and accurate information.
We use public posts only to build aggregate analytics and to train models for our own use. We do not sell, display, or redistribute anyone's posts without permission.
15+ years depending upon the age of the social network, community or forum.
~7,500 stocks (NYSE, NASDAQ, NYSE American), ~6,700 ETFs, and ~190 cryptocurrencies. More than 14,000 in total.
We have collected billions of conversations, and discarded billions more as spam, bots, or off-topic noise. Each post goes through several filters before reaching our dataset.
Collection runs continuously. Posts are classified and aggregated as they arrive. Timeseries data rolls up hourly.
Every post passes through two classifiers before entering the dataset. The first is a spam filter. The second is an asset-relevance filter ensuring that we are correctly differentiating between different assets which share the same names or tickers.
Each post is classified as Bullish, Bearish, or Neutral. We train separate models for each source and asset class because the language and norms differ substantially from one community to the next.
Each post is tagged with the topics it discusses. A separate binary classifier runs for each topic — a post is either about that topic or it isn't. Topics are defined per asset class:
Daily and monthly narrative summaries per asset. They cover sentiment shifts, active topics, and what's driving discussion.
Engagement is the total interaction count on a post — likes, replies, shares, upvotes, video views, whatever the source offers — summed into a single number. Video views count because watching a video is something a person chose to do. Impressions do not: we never count how many feeds a post happened to scroll past in. We aggregate engagement rather than post count, so a viral post with 10,000 shares is weighted more heavily than 100 posts nobody touched.
Attention is an asset's share of total engagement across all assets in its class, over a rolling window. If NVDA generates 3% of all stock engagement in a 24-hour window, its daily attention is 3%.
Net bullish = max(0, bullish engagement - bearish engagement), normalized against the total net-bullish conviction across all assets in the class — not total engagement. Assets with evenly split sentiment contribute nothing to that pool, so it's small and concentrated: a single asset can legitimately hold most of a day's net-bearish conviction. Net bearish is the inverse, and an asset is either net bullish or net bearish in a given window, never both. This is engagement-weighted — it reflects the sentiment of posts people actually interact with, not just post volume.
Twelve Data. Daily and hourly prices for stocks, ETFs, and cryptocurrencies.
Yes. Each stock's financials page covers earnings (actuals against analyst estimates), revenue, full financial statements, key statistics and ratios, and dividends. Fundamental data comes from Twelve Data, the same source as our price data.
Yes. Save assets into portfolios to follow them together.
Yes. Please view our dedicated API documentation.
Yes. Connect your favorite AI to our MCP server by following the connection instructions.