Sources & Importing

How Sources Work

Understand what Sources are, why they matter for AI enrichment, and how they differ from product imports.

On this page

How Sources Work

Sources are the reference material you give Merchkit to help the AI understand your products better. Think of them as context documents that the AI reads while enriching your product data.

What Sources Are (and What They're Not)

Sources are NOT product records. When you import a CSV of products or connect to an integration, that brings in your catalog data—SKUs, names, basic attributes. Sources are different. They're supplementary material: URLs, text snippets, documents that give the AI richer context about each product.

Here's the distinction:

  • Product imports (CSV, integration): Your core product data. SKUs, names, categories, prices.
  • Sources: Context material that supplements those records. Supplier spec sheets, product pages, technical documentation, descriptions.

When the AI generates attribute values for your products, it reads:

  1. The product data itself
  2. The sources you've attached to that product
  3. Your enrichment prompt

Together, these three inputs produce richer, more accurate attribute values.

Where You Interact with Sources

You interact with Sources in two places:

1. Per-Product Sources Panel

Open any product in your Products table and click its Data Sources cell. A panel appears showing all sources attached to that specific product. This is where you add sources one product at a time, or copy sources from parent products to variants.

2. Sources Sidebar Page

Navigate to Catalog → Sources in the sidebar. This shows a flat list of all sources across your entire catalog. Use this page to bulk-import sources via CSV, or to view and manage your source library at scale.

What Makes a Good Source

Strong sources:

  • Supplier specification sheets (PDFs with technical details, dimensions, materials)
  • Manufacturer product pages (official specs, features, certifications)
  • Data sheets with structured information
  • Vendor documentation (installation guides, compliance docs)
  • Internal product briefs or marketing materials you trust

Weak sources:

  • Marketing fluff without substance
  • Unrelated web pages
  • Generic category pages (not product-specific)
  • Social media posts or customer reviews (unless that's specifically what you want to enrich)

How Sources Feed AI Enrichment

When you run enrichment on a product with Sources attached:

  1. The AI reads the product record (name, basic attributes, etc.)
  2. The AI reads all sources attached to that product (URLs, text, documents)
  3. The AI reads your enrichment prompt (what attributes you want to generate)
  4. The AI synthesizes all three to produce attribute values

More sources = more context. More context = better, more confident enrichment.

Next Steps

Ready to add sources to your products? Move to Adding Sources to a Product for a step-by-step walkthrough.