How Prism actually scores a stock.

Prism is a derivative tool — every fundamental or price metric is sourced from a named third party, never invented internally. We deliberately keep the source list short so users can audit any number end-to-end.

Yahoo Finance fields are best-effort; thinly traded names may have null fields, in which case the affected rule reports as "missing data" rather than failing or passing silently.

Each preset framework is a list of rules of the form{ metricId, op, threshold, weight }. Evaluation runs the rule list against the stock's current metrics, computes a weighted pass-rate (0-100), and maps to a fit label using the framework's fitBands. The full rule list is inspectable in Studio and every threshold is user-adjustable.

Why this design. A framework is a published opinion of a real investor; we encode the most-cited screen they describe. For example, the Graham Number rule encodes the formula √(22.5 × EPS × BVPS)1 exactly as published. We do not try to capture qualitative judgement (moats, management quality, narrative) — those remain the user's job.

Three independent estimates are blended, weighted by quality:

• Earnings power — trailing EPS × quality-adjusted multiple. Multiple is 12× base, +0.4× per ROE point above 10%, +0.6× per growth point above 5%, capped 8×–35×.
• FCF yield — FCF/share ÷ a demanded yield. Demanded yield is 5% baseline, +1pp per 0.5× net-debt/EBITDA above 1, capped at 12%.
• Asset value — book value/share × an asset multiple from 1.0× (low ROE) up to 3.0× (25%+ ROE).

Compounders (ROE ≥ 15%) weight earnings + FCF at 0.45/0.45/0.10. Capital destroyers (ROE < 15%) flip the asset weight up to 0.40 so we don't capitalise optimistic earnings on a deteriorating book.

The margin-of-safety price defaults to 30% below the central estimate. The band visualisation shows current price relative to MoS / low / base / high.

Inverts the standard DCF: given price and a discount rate, what FCF growth must the market be assuming? Solved by bisection over [-15%, 60%] for 60 iterations. Discount is risk-adjusted (8% baseline + 1pp per net-debt/EBITDA above 1, capped 13%). Terminal growth is 2.5%, explicit projection 10 years.

Verdicts (conservative / reasonable / aggressive / extreme / unattainable) are calibrated against the company's own historical EPS+revenue blend, when available; absolute thresholds are used as fallback.

Seven dimensions, each scored 0-100 from a small fixed set of metrics, then mapped to a letter:

Overall is a weighted average — Quality 25%, Valuation 25%, Growth 15%, Balance Sheet 10%, Risk 15%, Momentum 5%, Income 5%. Categories with insufficient data are skipped and weights renormalised.

Letter cutoffs: A+ ≥ 95, A ≥ 88, A- ≥ 82, B+ ≥ 76, B ≥ 70, B- ≥ 64, C+ ≥ 58, C ≥ 52, C- ≥ 46, D ≥ 35, F otherwise.

Pattern-matches the current metrics against nine named failure modes: value trap, quality trap, yield trap, momentum trap, cyclical peak, leverage trap, growth deceleration, accounting red flag, dilution risk. Each pattern requires multiple corroborating signals — never a single metric — so flags are deliberately conservative.

Severity (Low / Medium / High) is determined by how many of the pattern's signals are present. Each flag exposes its evidence, what would reduce the risk, and what to monitor — so the user can disagree with the flag from a position of information.

Classifies the nature of the debate, not just polarisation. Each framework is assigned a style bucket (quality / value / deep-value / growth / momentum / income). The endorsing and rejecting camps' style sets feed a small rule-set that maps to named debate kinds: quality-vs-valuation, growth-vs-expectations, turnaround-vs-value-trap, momentum-vs-fundamentals, income-vs-sustainability, cyclical-timing, plus consensus shortcuts.

Polarisation is the standard deviation of all framework scores, scaled to 0-100 — a secondary signal under the headline classification.

Important things Prism makes no attempt to capture:

• Qualitative moats — switching costs, network effects, brand, regulation.
• Management quality — capital allocation track record, incentives, integrity.
• Industry structure — Porter's-five-forces dynamics, competitive intensity.
• Macro / cycles — interest-rate regime, currency, geopolitical risk.
• Path-dependent risks — pending litigation, FDA outcomes, single-customer concentration.

A clean Prism read is necessary but not sufficient for ownership. Treat the verdict as a checklist that the quantitative case is intact, not as an instruction.