Data snapshot model
DividendTen financial-data pages are dated snapshots, not live feeds. The market dataset has its own `updatedAt` date, which represents the underlying snapshot. A site deployment or editorial change does not alter that date.
The default freshness window is 45 days. When a verified, partially verified, or initial market snapshot exceeds that window, the application changes its effective presentation status to Historical snapshot. Seed, demo, and unresolved unknown-status datasets remain non-indexable rather than becoming publishable merely because they are old.
Dividend-yield fields
Yield is a ratio between dividend amounts included in a calculation and the price or benchmark input used for that snapshot. The result depends on the trailing period, price date, handling of special distributions, and source methodology.
DividendTen treats yield values as dated descriptive fields. They do not represent a forecast. When the dataset is historical, page titles and supporting text identify it as historical rather than using “right now” or similar live-market language.
Calendar fields
Calendar rows can include company name, ticker, ex-dividend date, record date, payment date, amount, currency, and payment-frequency label. All stored dates must be real calendar dates in ISO `YYYY-MM-DD` format.
The build validates that record dates do not precede ex-dates, payment dates do not precede record dates, and calendar rows remain sorted by ex-date. When every listed ex-date has passed, the page describes the table as historical instead of upcoming.
Payout-frequency classification
Frequency rows group the dated snapshot into quarterly, semi-annual, and annual or irregular categories. The build validates that percentage shares are finite and total approximately 100%.
Frequency is descriptive. It does not establish future payment behavior or quality, and a historical frequency table should be re-verified when a current classification is needed.
Source and provenance policy
Production financial data should be reconciled against primary company announcements, exchange/index information, or properly licensed market-data providers. Source and verification metadata must be recorded only when it actually exists.
The repository does not claim a blanket open-data license for the underlying market data. Dataset structured data therefore omits a license until a verified applicable license can be established.
Build-time quality checks
Validation covers required data fields, strict dates, duplicate tickers, rank order, descending yield order, percentage ranges, currency formatting, calendar chronology, semantic HTML tables, canonical hosts, structured data, internal links, indexability, and freshness.
`npm run check:freshness` warns when snapshots are outside the normal freshness window. `npm run check:freshness:strict` fails the release when a dataset is too old to be promoted as newly refreshed/current.
Material updates vs routine reviews
Article and sitemap modification dates should move only for material content changes. A routine editorial review may update a visible review date, but it must not automatically change sitemap `lastmod`. Market-data pages retain their real dataset date until the underlying financial snapshot is refreshed.