Where the information comes from
Research draws on company filings, technical publications, company disclosures, and attributed reporting. Use the source attached to a claim to check what it establishes and what it leaves unresolved. A document’s authority does not make every statement within it independently verified.
What the labels mean
| Label | Meaning | Reader check |
|---|---|---|
| FACT | A statement supported by the cited evidence within its stated scope | Does the source support the exact claim and date? |
| CLAIM | A company’s or other interested party’s assertion | Has anyone independently checked it? |
| PRESS | Attributed reporting | Is the underlying source available? |
| DERIVED | A calculation from stated inputs | Can the inputs and method reproduce it? |
| INFERENCE | An analytical interpretation | What assumptions or alternative explanations matter? |
| UNVERIFIED | An unresolved statement | What evidence is still needed? |
Repeated coverage of the same announcement is one underlying source, not several independent confirmations. An article that cites RoboIndex cannot independently corroborate RoboIndex’s own claim.
Confidence and dates
A confidence label expresses a research assessment. It is not a probability unless a separately documented, calibrated method defines it that way. A company can have strong evidence for revenue and weak evidence for a technical claim; use claim-level scope where available.
The publication date identifies when the source appeared. A verification date identifies when the record was checked. The research cut identifies the evidence window for a particular analysis. A website update does not refresh all three dates.
Missing information and comparisons
Unknown values remain unknown. A blank commercial-evidence field means that the structured assessment is incomplete, not that the underlying business activity is absent. Company claims, assumptions, and calculations stay distinguishable.
Compare operational metrics only when tasks, conditions, observation periods, and human assistance are compatible. Keep company-specific limits beside favorable results.
Corrections and methodology changes
Material corrections should identify the previous statement, corrected statement, supporting source, and affected research. Detailed change history is kept in the GitHub repository. Public methodology pages retain the rules and limitations needed to interpret the current output.
Evidence may prompt a review. Changes to scoring rules, scores, eligibility, constituents, weights, or research cuts require owner approval before publication. The CE-N review distinguishes proposals from the stored scores.
Portfolio research
RoboIndex presents research portfolio constructions. Historical simulations are backtests, not a live track record. Read the methodology for construction and performance limitations. Research use only; not investment advice.