Data Methodology

How We Collect Data

Every airline record on CarrySizer is sourced directly from official airline websites and their published fare rules. We do not rely on aggregators, crowdsourced submissions, or third-party data vendors for our primary figures. Each entry links back to the specific page or document we used so you can verify it yourself.

Primary sources include airline carry-on policy pages, fare conditions PDFs, checked-baggage fee schedules, and — where available — official customer service confirmation via the airline's help center. When official sources conflict (for example, a website says one dimension and a fare PDF says another), we use the more conservative figure and flag the discrepancy in our notes.

What We Check

For each airline and cabin class combination, we record:

• Carry-on dimensions— length, width, and depth in centimetres, plus the linear sum (L+W+D).
• Weight limit— maximum carry-on weight in kilograms where the airline specifies one.
• Personal item— whether a personal item (handbag, laptop bag, backpack) is permitted in addition to the carry-on, plus its size and weight limits.
• Gate-check fees— the fee charged when a carry-on is rejected at the gate and must be checked, where published.
• Cabin enforcement notes— whether the airline is known to enforce size limits strictly, spot-check only, or rarely enforce.

Quality Gate Criteria

An airline entry must pass our internal quality gate before it appears on the site. All four of the following must be true:

Freshness Levels

We classify each airline entry's freshness based on itslast_checkeddate:

Source Confidence

Freshness is only the age of a record. Confidence is stricter: a page is labelled High source confidence only when it has an explicitverified_atdate, an official-source tier, a clean source-diff status, and present expected markers such as dimensions or weight. Legacy batch-imported records are shown as source review pending until those checks are stored.

We keepsource_captured_atseparate fromverified_atso an automated fetch never pretends to be a human verification.