How StockMeter scores work
StockMeter publishes two scores. The Quant Score is a rules-based, objective summary of public data about a stock. The StockMeter Score starts from the Quant Score and adds a limited, AI-generated adjustment described below. The same scores are computed for every reader — they are never personalized and they are not investment advice.
What goes in
Each stock page shows five pillar grades — Quality, Growth, Valuation, Momentum, and Smart Money. The Quant Score combines four of them: Quality (gross profitability; return on capital), Valuation (earnings, free-cash-flow and operating-earnings yields relative to price or enterprise value), Momentum (12-month price trend, excluding the latest month), and Smart Money (open-market purchases by company insiders that are not part of a routine pattern, from Form 4 filings). These come from company filings (via SEC EDGAR, and standardized financial data derived from them), end-of-day prices, and insider ownership filings.
The Growth grade is shown for information, labelled “not in composite — under validation”; it is not part of the score.
Other inputs we research — for example institutional ownership (Form 13F) and activist ownership filings (Schedule 13D) — appear in the score only after validation and a changelog entry. Congressional trading disclosures and short interest are shown as information on the site but are not score inputs. Popular chart indicators (RSI, MACD, moving-average crosses) and the 30-week trend “stage” label are shown on charts as context — also not score inputs.
How the Quant Score is combined
Each input is compared with similar companies — the stock's industry, or its broader sector when the industry has too few companies (so banks are not compared with software companies) — and converted to a percentile rank. Inputs are averaged into pillars, and the four scored pillars are weighted equally, using fixed, published rules with no per-user tuning. Every Quant Score breaks down to pillars → factors → the underlying numbers, each traceable to its source filing or data field.
How the AI adjustment works (StockMeter Score)
A large language model (LLM) reads the company's recent SEC filings — including 8-K current reports and earnings releases — and the claims in the stock's StockMeter report, and adjusts the Quant Score by at most ±10 points. Every point of adjustment cites a specific source, and each claim is checked against that source before publication. The adjustment changes only when a report is generated or refreshed; between reports, the last verified adjustment carries forward (zero if the stock has no report).
AI-generated; may contain errors
Timing and honesty about data
Scores use only information that was publicly available on the score's date (“point-in-time”). Each day's score is stored permanently and never rewritten; later corrections appear as new entries. A stock needs data for at least three of the four scored pillars to receive a score. Some stocks carry thinner data — for example, foreign issuers and many ADRs do not file insider Form 4 reports, so their Smart Money pillar is left out — and are flagged accordingly.
Updates and versions
The Quant Score is recomputed once each night after the market close. When the methodology changes, the new version is documented in a public changelog with its effective date; historical scores keep the methodology version that produced them.
What this is not
Score bands (High / Mid / Low) are neutral summaries of data, not recommendations. StockMeter does not publish backtested or hypothetical performance claims.