Winning in English and losing in German.
European stores sell in three or four languages and score well in one of them. The German pages were translated once, two years ago, and the specification table, the FAQ and half the metafields never made the trip. An assistant answering a German query reads that page, not your English one.
Parity is measured field by field, against your source locale.
Substance, not word count
Whether the description carries the same facts, not whether it has the same number of characters. A short German page that states fit and materials beats a long one that does not.
Structure carried across
Identifiers, schema, language tags, specification metafields and FAQ blocks are checked per locale. These are deterministic and are carried across for free.
Answers per language
A buyer question answered in English only counts as unanswered in German. Warranty and returns questions are the ones that fail most often.
Where the German pages actually lose.
Not in the title, which was translated. In the specification table, the FAQ and the schema language tag, which were not.
Parity against the English source. Description substance EN 100, NL 88, DE 41. Specifications present EN 96, NL 74, DE 22. Identifiers carried EN 84, NL 84, DE 84. FAQ answered EN 61, NL 38, DE 9. Schema language tag EN 92, NL 55, DE 48. Locale parity comparison for the Nordvelo catalog. Schematic, not a screenshot.
Close it as a locale batch, once.
The German gap becomes one batch with one estimate and one approval: 240 pages brought level with their English source. Structural fields are carried across at no cost, prose is drafted from the source product data and reviewed, and a rule then requires the same fields in every locale so the next import does not reopen it.
Nordvelo: 1,400 products across 3 locales. English source 62, Dutch 55, German 38. Nordvelo · 1,400 products, 3 locales
Start with the score.
Free, about a minute, your domain and nothing else. Connecting your store is a separate decision you make afterwards.