A daily 0–100 score for how a management team is talking about the business. Comparable across peers, stable through history, and grounded in primary sources.
A Signal That Arrives Between Prints
Management language carries information the next print has not yet revealed. A team running a confident business talks differently from a team managing expectations down. The shift usually shows up one or two communications ahead of the metric, in the call, the press release, or the investor day deck.
Sentiment Analysis condenses every primary communication a company puts out into a single 0–100 score, updated daily, so the shift is something you watch rather than something you have to remember to listen for.
Live sentiment across four widely-covered names
Where the Score Earns Its Place
The score sits alongside revenue, cash flow, and valuation rather than replacing any of them. Numbers tell you what has already happened. Sentiment tells you how management is framing what comes next, and that framing usually moves first.
It works hardest as a screen. A name moving against its peer group, or against its own history, surfaces fast as something worth a closer look before the underlying numbers have caught up. From there the workflow is the same one you would already run, just pointed at a name you might otherwise have read past.
What a Divergence Looks Like in Practice
A semi name's score sits in the mid-70s for six quarters, broadly in line with the group. Through one earnings cycle, the call comes out and the score drops eight points. Two communications later (a Goldman conference appearance, then the next press release), it has slid to the high 50s. Peer scores have not moved.
Numbers from that company are still inside the range guidance allows. The 10-Q has not landed yet. What has happened is that management has stopped using the language they used before. Forward-looking statements have shortened. Q&A answers have hedged on the segment that was previously the bull-case driver. The score moved because the language moved.
A buy-side analyst who is watching the sentiment series asks two questions on the back of that drop. First, does the Common Ground view confirm the change in language and pin it to specific passages. Second, does Guidance Tracking show recent forecasts being met, missed, or quietly dropped. Both questions are answered inside the platform from the same primary documents that produced the score, with a citation for every claim.
The tone shift you would otherwise have caught two prints later shows up the day the call drops.
Start your free evaluation
Analyze 15 leading companies immediately. No registration, no credit card, no sales call.
Context-Aware, Not Word-Counting
Early sentiment tools counted positive and negative words against a fixed dictionary. Some pulled Twitter feeds. The category earned a reputation for noise because it missed the things language actually does: hedging, qualification, cautious optimism, the quiet pivot in how a CFO frames the quarter ahead.
Marvin reads each communication as a whole, weighing how words relate inside the sentence and how the sentence relates to everything the company has said before. Hedged confidence reads as hedged confidence, not as the average of "confident" and "uncertain".
Scores are normalized across industries and market phases. A 70 means the same thing for a software name as for a utility. That normalization is what makes peer comparison hold up: if a semi name sits at 75 and the rest of its group is at 68, the gap is a real divergence, not an artifact of how that sector talks about its business.
Only What Management Says
No financial media. No social media. No promotional press releases that were not filed with regulators. The score reads what management itself is communicating, not how a journalist or a posting account is reacting to it.
Inputs are 10-Ks and 10-Qs, prepared remarks and Q&A on earnings calls, regulatory press releases, investor day presentations, and the conferences where companies actually talk to investors: JPM Healthcare, Morgan Stanley Tech, Goldman Internet, WWDC, GTC, Build. Every source comes through the same Automated Data Import pipeline that powers the rest of the platform, ingested within seconds of publication.
