Localisation is a translation layer, not a market. That sounds obvious until you try to build a six-market panel and discover that a German-language page can be a British shop selling in pounds.
Beauty Bay's German page was the clearest case. The text was German, the currency was GBP, and the fulfilment was UK. Including it as a German observation would have imported a UK price into a euro median and a UK pack into a German sheet count. We left it out.
What we required instead
A record entered a country panel only when the page or a shipping policy established that country as the destination. Your Kaya's listing on ezebra established Germany explicitly, so it stayed. Two iHerb records for Clean Skin Club and Fomin entered the German panel through iHerb's German shipping policy, with the price left blank where the interface priced in the cart. The same brand family appears again in the Korean panel through iHerb Korea, with the destination established the same way.
Language was never used as evidence. That rule removed a page that looked German, and it kept pages that looked international but were explicitly shipping to the market.
What happens to a market total when you do this
The German panel reached twenty records with three identified cross-border entries. Korea reached twenty with three. The UAE reached twenty with eight purchase-agent entries. Strip those out and the numbers are Germany 17, Korea 17, UAE 12. Those three figures are the ones to quote if you want a local-channel view; the twenty-record figures are the panel as collected.
Neither set is a market size. They are descriptions of what a defined collection method returned.
What this does not say
We did not audit every page for fulfilment origin, and a page that ships locally may still import its stock. The point is narrower: the country label came from a stated destination, not from a language setting. Other Beauty Bay localisations or other regional storefronts may behave differently, and we make no claim about them.