Home/Slate Index Data Policy
Data governance

Slate Index Data Policy

The public index is built from public software-market evidence, versioned AI answer observations, and Slate-calculated ranking fields. Public pages do not use private customer workspace content as ranking evidence.

Content owner
Slate Index data governance team
Reviewed by
Phase 3 release approver
Last reviewed

Source classes

Observed AI evidence includes category buyer questions, platform/provider run labels, completed answer text that passes the public-answer gate, matched product mentions, positions, and cited domains. Product and vendor facts come from public vendor pages or other listed public sources. Slate-calculated fields include mention rate, position score, visibility score, rank, rollups, and publication eligibility.

Pages label or position these source classes so that an observed answer is not presented as a vendor fact and a Slate calculation is not presented as a direct quote. Source links and limitations remain close to the claims they qualify whenever the underlying public contract supplies them.

Collection and public/private separation

Ranking runs use versioned category prompts and published product rosters. The public catalog is designed to be reproducible from public sources and dedicated Brand Directory evidence. Private Slate customer workspace content is not used to rank products in the public index.

Separate optional forms may collect information a person chooses to submit, such as an email address, correction message, evidence URL, brand claim, watch request, or export request. That submitted personal information is handled under Slate's Privacy Notice and is not converted into ranking evidence merely because it was submitted.

Provenance and versions

Published ranking rows retain their snapshot and data window plus methodology, scoring, prompt-set, roster, and extraction versions. Prompt run references retain platform, provider, model, status, timing, and evidence locations. Public pages show the dates and evidence needed to interpret the current result without exposing private raw artifacts.

A newly collected answer does not silently rewrite an older promoted snapshot. Publication moves through validation and promotion, and pages read the latest eligible promoted evidence. This allows a failed or incomplete candidate run to remain separate from the current public result.

Freshness and retention

The scheduled ranking cadence is weekly with a rolling 30-day window by default, but freshness is governed by successful validation and promotion rather than a calendar promise. A page's visible date describes its applicable snapshot, evidence, enrichment, or reviewed policy revision.

Snapshot, source, audit, and correction records may be retained to support reproducibility, review, security, and rollback. Slate does not publish a fixed universal retention period for every evidence class. Personal-information retention and deletion requests are governed by the official Privacy Notice and applicable obligations.

Quality, exclusions, and unavailable values

Public gates exclude nonfinite or out-of-range metrics, reserved public slugs, unresolved workflow markers, stale entity relationships, ineligible answers, and pages without the required parent evidence. An unavailable metric is represented as unavailable or null rather than zero, NaN, or Infinity.

Slate self-cited runs are excluded from rankings. Public prompt answers require eligible external citation evidence. A product or page can therefore remain in source records while being withheld from public discovery until the applicable evidence and quality gates pass.

Corrections and material changes

Correction submissions create a review record and review task; they do not directly edit the public catalog. Reviewers compare the request with source evidence, then accept, reject, or request more information. An approved material correction updates the affected public record and its publication timestamp through the controlled publication path.