The shade names say "Light/Medium." The ingredient lists don't match the boxes. The returns page says "contact us."

The situation

What makes this catalog hard for AI to read

The brand sells skincare, haircare, and color cosmetics through its own Shopify store. Three product lines, each from a different supplier, each with its own naming convention. Shade variants are labeled "Light/Medium," "02," or "Warm Beige" depending on which line they belong to — sometimes within the same collection page.

Ingredient lists are copy-pasted from supplier Certificate of Analysis sheets. Some list INCI names, some use marketing names ("our proprietary hydration complex"), and some mix both. A customer asking "is this foundation vegan?" gets a different quality of answer depending on which product they're looking at.

The returns policy page exists, but it says "contact us for returns" and doesn't address whether opened cosmetics are eligible, what the window is for subscription orders, or how exchanges work for shade mismatches. When an AI shopping assistant encounters this, it either makes up an answer or says "I don't know" — neither of which helps the shopper buy.

The work

What Taz actually does with this catalog

Taz crawls every product page, collection page, FAQ, and policy page on the store. The audit is mechanical and exhaustive — not a sample.

• Variant audit: Every SKU is checked for shade-name consistency, weight/volume accuracy, and whether variant labels would make sense to someone who cannot see the product photo. "Size 1" becomes "30 mL Travel Size." "Light/Medium" becomes "Fair to Light with Warm Undertone."
• Ingredient reconciliation: Ingredient lists are compared against supplier sheets where provided. Marketing names are mapped to INCI names. Missing allergen callouts (gluten, fragrance, nut-derived) are flagged. Each product gets a structured ingredient record.
• Policy rewrite: Returns, shipping, and subscription terms are rewritten into structured Q&A pairs that answer the actual questions shoppers ask — "Can I return an opened serum?", "What happens if my subscription shade is discontinued?", "Do you ship to Canada?"
• FAQ gap analysis: Support tickets and review complaints (where provided) are scanned for recurring questions that have no published answer. Each gap becomes a drafted FAQ entry with a sourced answer.
• Agent answer pack: A structured JSON file with every product fact, policy answer, and FAQ entry — sourced only from crawled and verified data. This is what an AI assistant would use to answer shopper questions without hallucinating.

The deliverable

What the brand receives

• Product data audit spreadsheet — Every SKU in one sheet: current variant name, recommended name, ingredient completeness score, missing fields, and priority ranking by customer-facing impact.
• Rewritten FAQ — 15–25 Q&A pairs covering ingredients, allergens, returns, subscriptions, shipping, shade matching, and product care. Written to be extracted faithfully by AI search engines.
• Structured policy documents — Returns, shipping, and subscription terms in markdown with clear headings and conditional logic ("If your order is a subscription AND the product is opened, then...").
• Implementation ticket list — Prioritized CMS changes with exact copy for each. The brand's developer or Shopify admin can work through the list without needing to interpret anything.
• Agent answer pack (JSON) — Machine-readable product facts, policy answers, and FAQ entries. Every answer traces back to a source URL or document.