Every analyst agrees management quality matters. Most evaluate it on a handshake and a hunch. Marvin grades management on the public record: what they promised, whether they delivered, and whether the misses get explained or quietly dropped.
The problem with how management quality usually gets evaluated
The conventional inputs to a management quality view (the handshake, the cut of the suit, the glossiness of the IR deck, "how the CEO came across on the call") do not survive contact with a PM who pushes back. They are not auditable, not comparable across names, and not defensible six months later when the position has gone the wrong way.
The Marvin Labs management quality framework argues for a sharper input: track what management has actually committed to, in the public record, and grade them on delivery. That moves the question from soft impression to structured evidence. The tooling to run the workflow is already in the product. This page documents the shape end to end.
What generic AI tools miss
Pointing a large language model at an earnings call and asking "is this management credible" produces text that sounds confident and means nothing. The model has no record of what this management said last quarter, no link to whether the factory opening they promised actually happened, no view of whether the margin guidance from the capital markets day got quietly dropped two cycles later.
Management credibility is a longitudinal question. It only makes sense across a record of statements and outcomes, with each link auditable to source. Marvin builds and maintains that record automatically. The output is a structured scorecard, not a paragraph of generated commentary that nobody can pressure-test.
The four-stage workflow
Stage 1: Extract
Guidance Tracking captures every forward-looking statement management makes across earnings calls, press releases, investor day presentations, and filings. Metric-based targets (revenue growth, margins, production output, capex). Event-based commitments (product launches, factory openings, regulatory approvals, reorganisations, cost-reduction initiatives). The qualitative guidance that hides between the two. Each statement is tagged to the specific passage and date it came from.
Stage 2: Standardise
Vague phrasing is where most management commitment risk lives, because vague phrasing is the easiest to walk back from. "Low to mid single digits" gets normalised into a bounded numeric range with the assumption documented and the original wording preserved. Each statement is assigned a clear metric or event, a timeframe, a scope, and a target. The structure is what makes the scorecard comparable across companies and across time.
Stage 3: Evaluate
When the relevant period ends, each statement is automatically labelled: met, exceeded, missed, or dropped. Metric-based statements are checked against reported results. Event-based statements are validated against public disclosures and reputable reporting. Every evaluation links back to the original commitment and the outcome, so the logic stays auditable. A dropped commitment is treated as a distinct category from a miss, because management going quiet on a target is different information from management trying and falling short.
Stage 4: Compare
The point of the structure is what comes after a single name. Across a coverage set, the workflow produces a comparable record: which managements consistently deliver, which consistently miss, which drop commitments quietly when they're inconvenient, which explain the misses and which go silent. That is the raw material for management quality commentary, whether the analyst is writing it for internal thesis work, a sell-side note, or a client conversation.
Start your free evaluation
Analyze 15 leading companies immediately. No registration, no credit card, no sales call.
What the scorecard is, and what it isn't
The scorecard is an input. The track record tells the analyst whether management has historically delivered. It does not tell the analyst whether they will this quarter, or whether the misses were excusable, or how much weight the pattern should carry against the rest of the thesis. Those judgments stay with the analyst.
The scorecard is also not a black-box rating. The Marvin Labs messaging guide is deliberate about not collapsing management quality into a single number, because the reason a manager missed matters as much as the fact that they missed. The output here is a structured record of statements and outcomes, with every link auditable to source. Interpretation is the analyst's job.
Where this sits in the wider workflow
Management quality is one of several solution patterns the platform supports. The structured scorecard is updated automatically as part of the earnings season workflow, so the picture refreshes every cycle without separate work. On the buy-side, the scorecard turns "is this management credible on the margin path they promised" into a query against the record rather than a vibe check. Sentiment Analysis tracks how management communicates. Guidance Tracking measures whether their commitments deliver. Used together, they pick up patterns like rising tone with slipping delivery, which is usually more informative than either signal alone.
