The Data Stack: Three Layers Every Brand Needs

OpenAI's Instant Checkout—announced with Walmart, Target, and over a million Shopify merchants—has seen slower rollout than anticipated. The culprit isn't technology. It's data.

Consider what "inconsistent data" actually looks like: - Your "Navy Blue" exists as "Dark Blue" in one system, "Midnight" in another - Your "Medium" is "Size M" in one feed, "Med" in another - Your inventory updates hourly but your product feed updates daily—creating phantom stock

These aren't edge cases. They're endemic. And they determine whether agents can transact with your products.

Layer 1: Product Information (The Foundation)

Every product needs comprehensive, accurate, structured data covering:

Basic attributes:

{
  "id": "SKU-12345",
  "name": "Ultraboost Running Shoe",
  "brand": "Adidas",
  "price": { "amount": 180.00, "currency": "USD" },
  "availability": "in_stock",
  "inventory_count": 47
}

Specifications (dimensions, materials, technical details, compatibility):

{
  "weight_oz": 11.5,
  "drop_mm": 10,
  "arch_support": "neutral",
  "waterproof": false
}

Use cases (what problems it solves, who it's for):

{
  "ideal_for": ["road_running", "daily_training"],
  "gait_type": ["neutral", "mild_overpronation"],
  "experience_level": ["intermediate", "advanced"]
}

The standard for completeness is higher than most companies realize. You need data detailed enough that an agent can match your product to specific user needs without additional context.

Layer 2: Distribution (Getting Data to Agents)

Comprehensive data is useless if it doesn't reach agents.

Where your data needs to be: - Major commerce platforms (Amazon, Shopify ecosystem) - Product data aggregators - Industry-specific databases - The emerging commerce protocols: UCP (Google), ACP (OpenAI), Copilot Checkout (Microsoft)

If you're on Shopify, you're likely already enrolled or eligible for all three protocols. If not, implementation should be on your roadmap.

Critical rule: Consistency.

If your product data differs between sources—different specs, different pricing, conflicting information—agents get confused or lose trust. Data governance across channels becomes essential.

Layer 3: Real-Time Capabilities

Static data isn't enough. Agents expect current information.

What you need: - APIs that expose live inventory - Dynamic pricing feeds - Real-time fulfillment options - Integration with operational systems

"In stock" at query time may not mean "in stock" at purchase time. Without real-time feeds, agents work from stale data—and failed transactions destroy trust.

The Build Sequence

Months 1-2: Audit current product data. Identify gaps in attributes, specifications, use cases.

Months 2-3: Implement Schema.org markup. Get structured data on every product page.

Months 3-4: Build inventory/pricing APIs. Start with top 20% of products.

Months 4-6: Enroll in commerce protocols (UCP, ACP). Expand API coverage.

Ongoing: Monitor for data drift. Update attributes as products change.

The Payoff

Guess is using Microsoft's catalog enrichment agent to extract product attributes from images and structure them for AI. Their Head of Innovation says they can now "turn product details into meaningful insights that help shoppers discover styles in real time."

That's what catalog-for-agents looks like: not just describing products, but structuring data so AI can match products to specific customer needs.

Building this stack isn't glamorous. It's data engineering. But it's the foundation everything else rests on. Companies that neglect it will be invisible regardless of other investments.

Which layer is your weakest?