In the raw feed, each product carries its variants as a nested array - which is fine for code, but useless for filtering in a spreadsheet: you can’t sort by price or search a SKU if that data is buried three levels deep inside a JSON blob.

This tool’s default format flattens that: every variant becomes its own row, with the parent product’s title, vendor, type and tags copied onto each one, alongside that variant’s own price, SKU, option values and image. A product with four sizes turns into four ordinary rows you can sort and filter like anything else.

If all you need is images, there’s a dedicated images-only format that does the same flattening at the image level instead.

The same logic extends naturally to images: a product with six photos would otherwise bury them in a nested list too, which is exactly what that separate images-only format flattens on its own - one row per photo, with its position and alt text, instead of one row per variant.