rumor

    We measure
    investor attention.

    4.3B+
    posts collected
    14.4K
    assets tracked
    4
    platforms
    13+
    years of data

    We’re on a mission to collect and analyze every public conversation about investing. We’re not there yet. So far: 4.3 billion posts from Twitter, Reddit, StockTwits, and YouTube, covering roughly 15,000 stocks, funds, and cryptocurrencies, with history reaching back more than thirteen years.

    Rumor turns that archive into a screener. For every asset we measure how much attention it’s getting, how that attention is shifting, and how the conversation splits between bulls and bears — and we let you rank, filter, chart, and compare on those numbers the way you would on a P/E ratio. The point isn’t to surface what’s trending this afternoon; anyone can see that. The point is that thirteen years of history make it possible to ask whether today’s excitement is actually unusual.

    Some things we don’t do, on purpose. We don’t tell you what to buy or sell — no ratings, no targets, no score engineered to drive a trade. We don’t claim attention causes price moves; when attention precedes a move, “precedes” is the word we use. And our methods are documented in the FAQ, including the places where they fall short. Rumor measures the conversation. What you do with it is yours.

    Where the signal comes from

    Every day, millions of posts mentioning stocks, funds, and crypto land on X, Reddit, StockTwits, and Bluesky. We ingest them all — billions of posts across thousands of assets, continuously.

    Meaning, as a vector

    Every incoming post is embedded — turned into a numerical representation of its meaning. The same vectors that power sentiment and topic models also power the filters that come next.

    Less data, better signal

    Raw social feeds are noisy. Spam, bots, crosspost duplicates, and posts where the ticker means something else (AMD the chip, not the stock) all get dropped. Roughly a third of what comes in never makes it through.

    Sentiment, in three shades

    Our proprietary sentiment model is trained on financial conversations — not general social chatter. It splits posts into positive, negative, and neutral, separating the voices that matter from the herd. Neutral posts are discarded the same way spam is.

    Topics, not tickers

    A second proprietary model tags what each post is actually about: earnings, macro, regulation, analyst calls, and dozens more. The output is a topic fingerprint for every ticker — the conversations worth listening to, pulled clear of the noise.

    Ranked, then summarized

    What survives is ranked by engagement and assembled into per-asset summaries. The output is a portrait for every ticker — sentiment mix, topic mix, the posts that drove the day — updated continuously.

    See it in action

    Image credit: “Seeing New York (The Flatiron)” — Charles Dana Gibson

    rumor
    You're viewing limited data.
    Log inSign up

    Sign up

    Log inPricing