The data_quality_scheme table

What a data_quality_rating_native actually means: its scale, its direction, and where its caution threshold sits.

Fields#

field	type	required	meaning
`code`	text	yes	'ademe_dqr'
`name`	text	yes	The full name, as the organisation writes it.
`publisher_code`	text		NULL for a cross-publisher standard
`scale_min`	double precision	yes	Best end of the rating scale.
`scale_max`	double precision	yes	Worst end of the rating scale.
`lower_is_better`	boolean	yes	true = a LOWER rating is better (ADEME DQR), false = higher is better.
`caution_threshold`	double precision		The scheme's own published "treat with caution beyond this" threshold, in the scheme's direction. NULL when the publisher states none.
`axes`	jsonb		The axes the rating aggregates, as {axis_code: meaning}. Every key of emission_factor_value.data_quality_detail_native is one of them (a value carries only the axes its publisher scored) and every score sits on [scale_min, scale_max]; the audit gates both (dq_detail_axis_mismatch).
`description`	text	yes	Free prose describing this row.
`url`	text		Where to find it at the source.
`tier_band_high`	double precision		The cut a rating must meet to read 'high' in data_quality_normalized: <= when lower_is_better, >= otherwise. The three tier_band_* cuts are all set or all NULL, and all NULL means registered but unwired: the rating is stored and served verbatim and the tier never reads it (plan CLN-0014).
`tier_band_medium`	double precision		The cut a rating must meet to read 'medium' once it misses tier_band_high.
`tier_band_low`	double precision		The cut a rating must meet to read 'low'; one that meets none of the three reads 'very_low'.

Referenced by#

emission_factor_value.data_quality_scheme_native.

Notes#

A rating never travels without its scheme, because the number alone is not comparable: one publisher's 1-5 with lower-is-better is not another's. For cross-library comparison use data_quality_normalized, never the raw rating.