About & Methodology

Coverage

We cover all S&P 500 constituents (about 503 tickers, since a few companies have two share classes). The constituent list is refreshed weekly. Companies removed from the index stop updating; newly added companies are backfilled with five years of daily prices and their most recent quarterly reports.

Each fiscal quarter of a company is one period: the earnings press release (8-K) and the quarterly or annual report (10-Q / 10-K) for the same quarter are merged, and only one analysis is produced. Period names follow the company's own fiscal-year convention, with the quarter-end month shown alongside.

Data sources

Every financial figure comes from documents the company filed with the U.S. Securities and Exchange Commission. Market data comes from licensed data vendors. We do not use forums, social media or unverified rumors.

• Earnings releases, 10-Q / 10-K reports and 8-K current reports: SEC EDGAR
• Structured financials (revenue, income, balance sheet, segment revenue): the XBRL data embedded in the filings and EDGAR's company facts API
• Institutional holdings: Form 13F filings. Insider transactions: Form 4 filings by officers and directors
• Prices, volume and valuation metrics: Finnhub and Alpaca
• Earnings calendar and analyst consensus estimates: Finnhub
• Treasury yields: U.S. Department of the Treasury. FX: European Central Bank reference rates. News: public reporting aggregated by Finnhub, always linked to the original source

How an earnings analysis is produced

The text is written by a large language model, but the model does not decide the numbers. The pipeline has four steps, each with programmatic checks. If any step fails, the analysis is not published.

• Extract: the model pulls revenue, EPS, margins, guidance and segment revenue from the press release and the report, and must quote the source text for key claims.
• Verify: extracted figures are compared line by line with the XBRL data from the statutory filing. Differences above 2% are flagged and the filed figure wins. Every quote must be found verbatim in the document; quotes that cannot be found are dropped.
• Assemble: the scorecard (actual vs. consensus), the guidance table, the revenue flow chart and the balance sheet summary are built by code from the verified numbers, not by the model. Both sides of the revenue flow chart must add up to total revenue.
• Write: the model only turns these fixed numbers into plain-language prose and summarizes management's commentary and key risks. It never gives buy or sell advice.

Update schedule

Around a company's earnings date we check for new filings every 30 minutes; an analysis is usually available about an hour after the press release is filed. When the 10-Q arrives days later, the structured figures are completed and re-verified with its XBRL data.

The Daily Wrap is generated each trading day in the evening (U.S. Eastern time) using closing data. Prices, valuation metrics and the earnings calendar update daily; institutional holdings update quarterly as 13Fs are filed. All market data on the site is end-of-day, not real time.

What the “FinDog View” is

The Bullish / Neutral / Bearish label on stock pages is a one-line judgment produced by a model from public information: recent revenue and earnings trends, valuation relative to the company's own history, whether the latest report beat expectations, guidance changes and recent news. It updates once per trading day and always comes with a reason.

It is a mechanical summary of public information. It is not investment advice or a recommendation to buy or sell, and it knows nothing about your finances, holdings or risk tolerance.

Chinese and English

The Chinese and English editions are written separately from the same verified numbers; neither is a translation of the other. If an analysis has not been generated in your language yet, opening the page while signed in starts generation, which usually takes a minute or two.

Limitations and known issues

We work hard on accuracy, but an automated pipeline will make mistakes. Known limitations:

• The model can misread source text, especially in industries with unusual reporting (banks, insurers, utilities, REITs). Some items, such as gross margin or the revenue flow chart, may be missing for those companies.
• Consensus estimates come from a third party and may differ slightly from figures you see elsewhere.
• If a company later corrects or restates a report, an existing analysis is not updated automatically.
• Earnings call transcripts are not covered yet; management commentary is taken from the MD&A section of the filings.

Corrections

If you find a wrong number or statement, email i@elsetech.app with the page URL and the source you believe is correct. Once confirmed, we will fix it and regenerate the page.

Disclaimer

Everything on FinDog is for information and investor education only. It is not investment advice, an offer or a solicitation. Investing involves risk; consider your own situation and consult a licensed professional where appropriate.