Capture · Product Catalog
The product data foundation your AI was missing.
Personalization stops at "who" without a model of "what." Flash adds an AI-ready catalog that flexes to your vertical— serialized luxury pieces, apparel size matrices, F&B combos — synced from your ERP and Shopify, and wired to every order line so your segments and agents finally reason about products, categories, and brands.
The problem
You know your members. You can't personalize on what they buy.
You can’t tell what they’ll buy next
You know who your members are — but not what to put in front of them. With no product, category, or brand affinity on the profile, next-best-product, cross-sell, and replenishment timing are guesswork.
One model can’t fit every vertical
Jewelry needs per-item serials, apparel needs size matrices, F&B needs combos. Generic catalogs force you to bend your business to the tool.
Your AI has nothing to read
A consumer bot can’t answer “is this gluten-free?” or “what size should I get?” with no specs, no size guides, no grounding content per product.
Industry models
One catalog that flexes to how your industry really sells.
Not a separate schema per vertical — a small set of structural primitives plus extensible attributes, so a new industry needs no migration. Each model below maps to a shipped table or field.
One-item-one-code (serialized units)
Every physical piece is tracked individually below the SKU — its own serial, authenticity certificate, and unit attributes (carat, metal, movement), with ownership recorded at sale.
Size / color variant matrix
Named option axes (Size, Color) with a sellable variant per combination — the model fit and return profile your stylists and AI need to recommend the right size.
Combos & modifier groups
Build-your-own meals: a parent product composed of required components and optional modifier groups (choose-a-side, add-a-drink) with per-option upcharges.
Brand · category · price lists
Hierarchical categories, brands, tiered/regional price lists, inventory by store or warehouse — the breadth a multi-category catalog needs.
AI infrastructure
Built so your agents and bots can reason about products.
14-table PIM
Products, variants, serialized items, components, media, content, inventory, price lists, collections, and normalized order line items — one canonical, team-scoped model.
External-mastered sync
ERP, Shopify, or any middleware pushes by its own external_id; Flash upserts idempotently and never writes back. One generic import endpoint.
AI grounding content
Every product carries rich descriptions, specs, FAQs, size & care guides, and certificates — the corpus a consumer AI bot reads to answer product questions. Browse and read it through MCP tools (list_products / get_product) or the in-app catalog console.
Category-level metrics & personalization
Order line items resolve to product, category, and brand — powering GMV-by-category/brand and top-product metrics in the governed semantic layer, plus affinity tags (multi-category shopper, purchased-category) for segments and AI.
Next-best-product recommendations
A co-purchase engine ("members who bought X also bought Y") plus category affinity produces per-member next-best-product picks — surfaced to campaigns as email merge data, to AI agents via the recommend_products MCP tool, and over the v2 API.
External-mastered sync
Your ERP stays the source of truth.
You don't re-key products into yet another system. ERP, Shopify, or any middleware pushes products, variants, categories, brands, and inventory keyed by its own ID; Flash upserts idempotently and never writes back. Shopify products map in automatically — images and rich descriptions included.
One generic import endpoint
POST /api/v2/catalog/import
{ "source": "erp",
"brands": [...], "categories": [...],
"products": [...], "variants": [...],
"inventory": [...] }Idempotent by (team, source, external_id). A bad row is reported in errors[] — it never aborts the batch. Read back via GET /api/v2/products,/categories,/brands.