Individends separates source-model values, derived presentation scores, evidence confidence and data-quality eligibility. They answer different questions and must never be treated as interchangeable.
The recorded DSS, DGPS and Dividend Score use a 0–10 display scale. The public Dividend Safety number is DSS multiplied by ten. A numeric scale does not establish a probability, a forecast or a shared model formula.
DSS is the source model’s estimate of the company’s capacity to sustain its current dividend. Individends does not recalculate or overwrite the shared source value. The site reads it, dates it and stores selected product-side snapshots for change monitoring.
DGPS is a source-model growth-potential signal. The connected records do not currently identify their producer version. Individends therefore shows the recorded value without assigning a decrease, stable or increase forecast. A calculator file’s version alone cannot establish which model produced an older stored observation.
Stock pages, watchlists and rankings retain the same stored source score. A separately dated ratios-table yield is displayed as a fact and does not recalculate that score. The deployed calculator and the local development calculator have different formulas; no formula is attributed to a stored observation until its producer version is established.
Recovery days is the source model’s estimate of how long the market price took, or may take, to recover the cash value removed by a dividend event. It is shown with its source date and never treated as a guaranteed future timetable.
Confidence is not the probability that a dividend will be paid. It is a quality score for the evidence currently available on the page.
A missing or implausible yield caps confidence at 25%. Missing events and aging source data reduce the score even when every database field is non-null.
Critical missing data caps this measure below the usable range. Financial inputs that have not been reconciled cannot receive a high-confidence label. Completeness and plausibility alone do not establish economic consistency.
All identified dividend listings in connected sources remain searchable and browsable, including stale and incomplete records. Catalogue inclusion is separate from whether a figure or model score is usable. Scored comparisons describe their evidence requirements; missing analysis never removes the underlying stock.
The headline TTM yield comes from the source ratios record and carries its own date. Ambiguous ticker symbols are not joined across exchanges. Calendar-year dividend totals are built from the available ex-date history, with the current year labelled YTD. Special dividends and incomplete histories can affect totals. The following quality checks control eligibility for a complete indexable research page, while partial-data stock pages remain discoverable:
Corporate actions, special dividends, currency units, stale prices and incomplete event records can distort dividend statistics. Individends suppresses suspicious values, but no automated policy can remove every ambiguity. “Review required” means the page should not be used as decision evidence until refreshed or reconciled.
Quality thresholds and confidence weights are operational checks, not calibrated payment probabilities. Model history coverage describes source history, not the chance that a dividend will be paid. The deployed source-model version is not inferred from a score alone.
The local July 2026 development study of Oslo and Stockholm stocks reported stored DSS test AUC of 0.4844 and stored DGPS balanced accuracy of 33.77%. The report states that its test period had been inspected previously. These results do not establish reliable forecasting across other markets or certify the currently deployed model. Broader and prospective validation remains required.
Individends provides research and decision support, not personalized investment advice. Scores describe current evidence under a documented method; they do not promise future returns or payments.