How a seat map gets made

The source mix varies by configuration: airline material, independent aviation references, and structured extraction can all appear in a map's record. A person reviews every map before publication. The confidence label is separate: it summarises the sources recorded for that map. Many maps cross-reference two or more sources; the label tells you which. A published verification record can also show when a check happened and who recorded it.

Here is how we record the evidence and checks that are available for a map.

1. We record the source

Source records can point to an airline diagram or another aviation reference. A source can support cabin splits, row numbers, columns, and facility positions, but its value depends on the configuration and date it describes. Missing evidence stays missing rather than being filled with a universal provenance claim.

2. Then we record the source mix

For the confidence label, repeated entries from the same source count once. The editor's own sign-off does not count as a source. A configuration is labelled Cross-referenced only when at least two distinct, non-human source names are on file and the map has a human review stamp. The rule counts source names; it does not decide that those sources agree on every seat, facility, or measurement.

3. We record the checks performed

Machine reading can turn a cabin diagram into structured data, but a transcription is not proof. Editorial and automated checks are recorded when they happen. Those checks can catch a shortened cabin, an invented row, or a column read from habit rather than the source. The public claim follows the record for that map, not a fleet-wide promise.

4. A seat count outside the expected range is refused

Before a map can go live, its seat total and its largest cabin are checked against the range we hold for that aircraft type. Both ranges are ours: the total is the span of commercial layouts we judge plausible for the airframe, and the largest-cabin floor is worked out from the configurations already in our database. A count outside the range is refused; there is no override that waves a map through with an explanation attached. We have not seeded a range for every aircraft type yet, so a type without one skips this check. The ranges are deliberately wide, so what this catches is gross truncation: a dropped cabin, or a total that could not belong to that aircraft. It will not catch a drift of a row or two, which is why source and editorial checks still matter.

5. Read the map's record

Where a published map has a verification record, it can name the editor, date, and checks performed. That record does not set the confidence label. The current confidence rule is:

• Single source: the default when fewer than two distinct, non-human source names are on file, source data is missing, or there is no human review stamp.
• Cross-referenced: at least two distinct, non-human source names are on file and the map has a human review stamp.
• Verified: the current derivation rule does not award this label. A legacy stored value can still display it.

Source names are compared without regard to capitalisation or surrounding spaces. Two diagrams from the same provider still count as one source.

What we add that the airline map doesn't

An airline diagram shows you where the seats are. It rarely tells you which ones are worth picking. After reviewing a layout, we can add practical notes about legroom and recline, window alignment, galley and lavatory noise, fast-exit rows, and quiet zones where the evidence supports them.

When a cabin changes

Airlines refit aircraft and swap configurations without much warning. When a change is confirmed by new evidence, we re-check the affected maps and move the verification date forward. Until we can confirm it, the existing date stays honest about when we last saw proof.

Found something wrong?

We would rather hear it. Every seat map page has a “Request Configuration” link for aircraft we have not covered yet, and you can flag a seat note that looks off at corrections@seatmap.app. We work through them in order of how many people flag the same map.