FIELD NOTES

One SKU, two warehouses. Is it really a duplicate?

Catalog Check field notes · Updated October 26, 2026 · Original synthetic examples

A duplicate identifier is a question about the data model. It is not an instruction to delete a row.

A SKU identifies a product in many catalogs. In an inventory export, however, one row may mean a product at a particular warehouse. The same SKU can legitimately appear more than once.

sku,warehouse,stock
001,A,5
001,B,8
002,A,3

There are three records here. The two 001 rows describe different warehouse stock positions. Choosing only sku would leave two possible records for the same identifier.

Row order can invent a change

Suppose the latest export puts warehouse B first and also changes its stock from 8 to 9:

sku,warehouse,stock
001,B,9
001,A,5
003,A,2

If a comparison pairs the first 001 with the first 001, it could report warehouse A becoming B and stock 5 becoming 9. That combines an order change with a real stock change and misses the business relationship.

Use the combination that identifies the record

For this example, select sku and warehouse as the matching key, then compare stock. The correct result is one changed record, one added record, one missing record and one unchanged record. Only warehouse B's stock changed among the matched records.

Catalog Check pauses on duplicate or partly empty keys instead of pairing them in input order. When a recognized warehouse or variant column would make the combination unique, it offers that column as a suggestion. You still need to confirm that it is a stable identifier.

A unique column is not always a suitable key. Stock and price can happen to be unique today, but they are values you expect to change. Adding them to a key can turn ordinary updates into apparently missing and added records.

Four checks before you choose another key

What does one row represent: a product, variant, location or something else?
Does the chosen combination identify one record in both exports?
Can any key component change as part of the update you are trying to detect?
Do both exports cover the same locations, status filters and time period?

When adding a column does not help

Repeated SKU-and-warehouse combinations may indicate repeated export rows, multiple bins, multiple stock states or another missing dimension. Examine the source system's export definition. Do not choose an arbitrary extra column just to force uniqueness.

If you ignore spaces or letter case in identifiers, distinct raw values can also collapse into one key. For example, A1 and A1 become the same after trimming. The tool blocks these collisions too, rather than quietly choosing a row.

Use the Multi-warehouse sample in the workbench to see the pause, inspect the conflicting rows and add the warehouse key. Then check the single stock change against the source examples above.

Try the multi-warehouse sample →