Structuring product data for multiple markets
Internationalization multiplies your catalog by your markets - unless you architect against it. The discipline that prevents multiplication: classifying every piece of product data as universal or market-dependent, and giving each class the right home. Get this right and adding a market is a projection; get it wrong and every market is a fork.
The classification pass
Walk your attribute model (Phase 2) with one question per field: does this fact change across markets? Universal: what the product physically is - materials, construction, measurements-as-made, GTINs, images (mostly), the option structure. These live once, in base fields and metafields, and are never duplicated per market. Translatable: same fact, different words - titles, descriptions, attribute display values ("Linen"/"Hør"/"Leinen"), care instructions, size guides' prose. These live once in the base language, with translations layered on top (next lesson's machinery). Genuinely market-dependent: facts that differ, not just translate - pricing (Markets handles natively), size conventions (an EU 38 presented as UK 10/US 6 - a conversion, not a translation), compliance-driven content (ingredient disclosure rules, labeling requirements per market - the Phase 11 beauty headache previews itself here), market-specific shipping/returns messaging, and occasionally market-specific merchandising (a "Midsummer" collection that only makes sense in the Nordics).
Where market-dependent data lives
The architecture principle: market-dependence should live in systems designed for it, not in duplicated products. Pricing: Markets' native per-market pricing. Display and content variation: translated metafields (Shopify's translation layer covers metafields - underused and exactly right for per-market attribute display) and market-conditional theme rendering. Size conversions: a lookup structure - one metafield holding the universal size, conversion tables rendering per market (and feeding per-market feeds, Phase 7's lookup-table pattern) - never twelve variant titles with hardcoded regional sizes. Inclusion/exclusion: Markets' publishing controls, driven by rules where possible (a "restricted-markets" metafield beats tribal knowledge about which products can't ship where). The anti-pattern to hunt down in existing setups: duplicated products per market ("Linen Shirt DE") - every one is a fork paying the synchronization tax forever, and each is usually a workaround for not knowing the mechanisms above exist.
The payoff structure
Notice what this architecture produces: the universal layer is enriched once and every market inherits it (Phase 2's work, now paying in every language); the translatable layer scales with languages, not markets (five markets sharing German cost one German translation); only the genuinely market-dependent layer scales with markets - and it's small when correctly classified. That cost structure is the whole game: brands that classify well expand into market eight for roughly the cost of market three; brands that fork catalogs find each market more expensive than the last, and usually can't say why. The audit worth running before your next market: sample twenty products, list every per-market difference that exists today, and ask of each - universal, translatable, or truly dependent? Whatever's misfiled is your expansion tax, itemized.