Multibrand: inheriting terrible product data from 40 different brand suppliers
The multibrand retailer's defining headache is that someone else makes most of your product data - and they make it badly, inconsistently, and differently per brand. Forty suppliers means forty spreadsheet formats, forty naming philosophies ("Navy" / "Marine" / "Dark Blue 019"), forty opinions on what a size run is called, image standards from studio-grade to phone-photo, and descriptions ranging from decent to absent to identical-across-every-stockist (Phase 4's duplicate-content trap, delivered to your inbox weekly). The retailer who pastes this chaos straight into the catalog operates a store with forty data models at once - which is to say, none.
The intake pipeline: your model at the door
The architecture is Phase 10's supplier gate, promoted from a paragraph to your core operating system. The principle: supplier data never touches the live catalog directly - it passes through an intake layer where your data model (Phase 2) is enforced. Per supplier, a maintained mapping recipe: their columns → your fields, their value vocabularies → your closed lists (their "Marine" → your "Navy", their size-run notation → your structured sizes - lookup tables, Phase 7's pattern pointing inward), their category labels → your taxonomy. Plus per-supplier validation at intake (Phase 10's rules, run before import): required fields present, GTINs valid, images meeting floor standards - with failures rejected back to a gap list, not imported as holes. The economics that justify the build: recipe construction is per-supplier one-time work (an hour to an afternoon each); without it, the same normalization happens implicitly, product-by-product, forever, by whoever uploads - the most expensive possible way to do it. New-supplier onboarding then gets a data clause: your import template and requirements sheet in the buying agreement - suppliers comply with surprising willingness when asked at contract time and near-zero willingness when asked per-upload.
The enrichment triage: where inherited data ends and your value begins
Intake normalizes; it can't create what suppliers never sent. The multibrand enrichment posture (Phase 10's loop with retail economics): triage by the standard exposure math - revenue products and filter-load-bearing attributes first - but with the multibrand-specific twist: your enrichment is your differentiation. Every stockist of a brand received the same thin description; the retailer who enriches it - real fit notes, materials structured, styling context, your photography where feasible - is the one whose page isn't a duplicate (Phase 4's economics: unique data is the entry fee for indexation, and in multibrand almost nobody pays it, which is precisely the opportunity). The triage output is also supplier feedback with teeth: per-brand data-quality scores from your validation layer, shared at buying meetings - "your products underperform on our filters because 60% lack material data" is a conversation that improves next season's intake, and buying teams hold leverage that data teams don't.
The coherence layer: one store, forty brands
The last mile is making forty brands feel like one store - which is entirely a consistency achievement: one taxonomy (products live in your categories, brand-agnostic - the customer shopping "linen shirts" doesn't care whose spreadsheet they came from), one attribute vocabulary (filters spanning brands only work because "Navy" means one thing catalog-wide - the whole point of the intake mapping), one image and title convention (title patterns normalized at intake: Brand + type + attributes, whatever your Phase 2 convention says - not forty formats), and brand as an attribute among attributes (a structured, filterable field feeding brand pages - which sets up the next lesson's fight). The compounding result: the intake pipeline, run for two years, produces a catalog more coherent than most single-brand stores - because you were forced to build the discipline they never needed. The headache, once more, inverted into the moat.