Editorial standards

Official sources first

Core specifications come from official manufacturer sources wherever a product has them: spec pages, manuals and PDFs, datasheets, and support docs. This is the gold standard and the source of the great majority of the index. Where a field is not documented by the manufacturer, it is stored as a null and said plainly, never filled from a weaker source.

A product that is sold only through retailers and has no official manufacturer page may still be indexed, from a retailer or marketplace listing, so the index can cover the whole market. These records are marked with a lower source tier, their retailer-sourced fields carry a lower confidence grade, and the page shows a sourcing note. An official manufacturer figure is always preferred and is never overridden by a retailer value.

Field-level provenance

Every core field carries its own provenance: the normalized value, the exact raw claim text it came from, the official source URL and type, the date checked, and a confidence grade. Conflicts between official sources are preserved and surfaced, not silently resolved. This is visible on entity pages and included in the JSON exports.

Honest nulls and anti-hype

• No guessing. Unknown specs are null, not inferred.
• No marketing inflation. Boost/surge modes never count as continuous output; partial-recharge claims never become full-recharge times; "up to" figures are treated as maxima.
• No unsupported real-world claims. No runtime hours, "days off-grid", or air-conditioner support without a stated load and assumption basis.
• Safety fields are stricter still. Generator CO-shutdown status is never inferred; ambiguous safety documentation blocks publication.

Cross-category honesty

Use-case guides compare product families (battery, generator, solar) on shared user-facing axes only — indoor safety, fumes, noise, output, runtime model, setup. They never merge the categories into one fake spec table, never equate watt-hours with gallons, and never hide a major tradeoff for a cleaner ranking. Recommendations are grouped by category with an explicit tradeoff summary, and the visible HTML always matches the machine-readable JSON.

Merchant independence

The index is merchant-agnostic. Product pages may list several merchants, and some of those outbound links are affiliate links — labelled affiliate in the table and marked rel="sponsored nofollow" in the markup. Merchant relationships have no input into which products are indexed, their specifications, rankings, verdicts, or ordering. Data and monetization are separate layers by design; the full terms are on the affiliate disclosure page.

Prices, where shown, are recorded by hand from the merchant's own page with the date they were checked, and are withheld once they age past the published freshness window rather than shown stale. No price is emitted in structured data unless the same figure is visible on the page. A merchant offer is never a source for a normalized specification; where a product is indexed from a retailer listing for lack of any official page, that sourcing is recorded as field-level provenance and marked with a retailer source tier, kept entirely separate from the commerce layer.

Corrections and freshness

Incorrect data is a defect. When an error is found, the entity is corrected and its updated_at date advances. Every entity shows a last-updated date on the page and in its machine summary, and the sitemap carries per-URL lastmod values.