Shopify SKU coverage on variants

A SKU is your private internal identifier for a specific variant — the medium navy crew-neck, not the t-shirt product page. GTIN identifies the same product across the world; SKU identifies the same variant inside your stack. AI shopping agents care about SKUs because once they have decided to surface your product, the user's follow-up actions — re-ordering, caching, "buy the same one again" — all need a stable handle to that exact variant. A blank or duplicate SKU forces the agent to guess; an agent that has to guess across catalogs decides "low confidence, skip." Shopify exposes SKUs at variants[].sku in the public products feed, and the bulk-fill is one of the cheapest score moves a Shopify operator can make.

Last updated 2026-04-30 · Deep signal · 6 pts

What it is

SKU stands for "stock-keeping unit." Inside Shopify, it's a free-text field on every variant. You set it; Shopify never auto-generates one (unlike GTIN, where Shopify will accept the manufacturer-issued UPC/EAN and validate the check digit). A typical SKU looks like TEE-NAVY-M, SHOE-BLK-09-W, or your ERP's internal ID like SKU48291. The encoding doesn't matter to AI agents — they just need a stable identifier. What matters is that the same variant always returns the same SKU and that no two variants share one.

"variants": [
  { "id": 1, "title": "Navy / S",
    "sku": "TEE-NAVY-S" },
  { "id": 2, "title": "Navy / M",
    "sku": "TEE-NAVY-M" },
  { "id": 3, "title": "Navy / L",
    "sku": "TEE-NAVY-L" }
]

"variants": [
  { "id": 1, "title": "Navy / S",
    "sku": "" },
  { "id": 2, "title": "Navy / M",
    "sku": null },
  { "id": 3, "title": "Navy / L",
    "sku": "" }
]

"variants": [
  { "id": 1, "title": "Navy / S",
    "sku": "TEE-NAVY" },
  { "id": 2, "title": "Navy / M",
    "sku": "TEE-NAVY" },
  { "id": 3, "title": "Navy / L",
    "sku": "TEE-NAVY" }
]

"variants": [
  { "id": 1, "title": "Navy / S",
    "sku": "auto-7f3a91b" },
  { "id": 2, "title": "Navy / M",
    "sku": "auto-2c4d8e1" },
  { "id": 3, "title": "Navy / L",
    "sku": "auto-9b6f3d2" }
]
// ...regenerated every PIM sync

The third shape — every variant getting the same SKU — passes a naive "is the field populated" check but fails the harder uniqueness test. Agents need to differentiate the medium from the small, and they cannot if both report TEE-NAVY. We surface this collision case as a separate "SKU duplicates detected" alert in the scan output even when the static signal scores half-credit.

SKU vs GTIN vs MPN: which is which?

SKU and GTIN are independent — set both. SKU is the cheapest to fix; GTIN often costs $30–$1,000 if you don't have a manufacturer-issued one (see GTIN coverage). For the rare case where you sell a third-party brand and the manufacturer's GTIN is the only stable cross-retailer identifier you have, set the GTIN as both the GTIN and the SKU; it's not ideal but it beats blank.

Why AI shopping agents care

• Stable handle for re-order intent. "Buy me the same medium navy tee I got last year" is exactly the kind of follow-up an AI shopping agent is supposed to handle. Without a per-variant SKU, the agent's only handle is the variant title string ("Navy / M") which collides across stores and changes when you rename a colorway. With a SKU, the agent has a value it can cache locally and look up against your feed deterministically.
• Cross-channel reconciliation. Your Shopify catalog feeds Google Merchant, Meta product feeds, Pinterest catalog, TikTok Shop, and the Shopify Global Catalog. Each surface uses SKU as the join key. Variants without SKUs fall out of channel sync silently — they ingest as a different product on each surface, fragmenting the listing.
• Inventory truth across stores. Multi-store brands (US site + UK site + B2B portal) reconcile inventory by SKU. Agents that observe the same product across multiple Shopify storefronts use the SKU to detect "this is the same variant, just in a different region" — and route the user to their region's store. Without SKUs, the variants look unrelated.
• Confidence in tied rankings. Catalog hygiene is a positive prior on tied rankings. SKU coverage is the easiest hygiene signal to read because it's a uniform per-variant field — a 100% covered catalog reads as well-managed; a 30% covered catalog reads as a side hustle.

How to test it on your store

One jq one-liner against your public products feed gives you both the coverage rate and the duplicate count.

curl -s 'https://yourstore.com/products.json?limit=250' \
  | jq '
      .products[].variants[].sku as $s |
      [$s] | length as $t |
      [$s | select(. != null and . != "")] | length as $f |
      "\($f) of \($t) variants have a SKU (\(($f*100/$t) | floor)%)"
    '

# Duplicate check
curl -s 'https://yourstore.com/products.json?limit=250' \
  | jq '[.products[].variants[].sku | select(. != null and . != "")] | group_by(.) | map(select(length > 1)) | length'

The first command returns your coverage percentage. Target ≥90%. The second returns the number of SKU values that appear on more than one variant — should be zero. Anything above zero means you have collisions, which we flag as a separate alert.

For stores with more than 250 products, run with ?page=2, ?page=3, etc. and concatenate. The CatalogScan free scan does this paginated walk for you up to the first 1,000 products and surfaces the per-page SKU coverage in case you have a problem isolated to one collection.

How to fix it

// Pseudo-code
for await (const product of paginate('products')) {
  for (const variant of product.variants) {
    if (!variant.sku) {
      const sku = encodeSku(product.handle, variant.option1, variant.option2);
      await mutate('productVariantUpdate', { id: variant.id, sku });
    }
  }
}

5 mistakes themes and operators keep making