OUR METHOD

We use AI-assisted curation and automated source checks. Eligible curator-selected records can publish automatically; exceptions go to editorial review. Listings are source research, not hands-on reviews.

Evidence, in plain sight.

There is no universal best bag. We keep measurements, experiences and estimates distinct so you can decide what matters.

Every claim has a trail

A specification is attached to its exact variant, source URL, excerpt and retrieval time. Automated extraction creates a candidate claim. Source validation or an editor checks it before publication. Conflicting claims are retained, including rejected alternatives, in the editorial record.

Reviews are not interchangeable

We distinguish mentions, owner reports, hands-on tests and long-term use. Independence and exact variant match are checked separately. A search snippet is only a discovery lead and never automatically becomes verified review evidence.

Geometry is a useful estimate

Bounding-box volume is width × height × depth, divided by 1,000 for millilitres. It is not usable capacity. Claimed-volume ratio compares stated capacity with that box. Weight per litre uses the stated capacity, not the bounding box.

Our descriptive thresholds are editorial rules: depth up to 30 mm is “very flat”, up to 60 mm is “slim”, and greater depth is “deep”. Bulky-object tolerance describes depth only. It is not a tested packing result.

Availability is regional and dated

Japan listings are explicit offers for JP. Yen pricing or a Japanese brand alone does not prove availability in Japan. Unknown means no approved evidence. Stock can change after a check.

Commercial independence

Commercial links must not change evidence classifications. Affiliate links, when configured, are labelled beside seller links. Commission does not affect selection or evidence. Advertising awaits publisher approval and consent setup.