The two metafield import footguns — and how to not fire them.

Footgun 1 — a blank cell means “delete”, not “skip”

This is the one that costs people a weekend. You export 4,000 metafield rows, you edit the eleven you cared about, and you re-import the whole sheet. The 3,989 rows you didn't touch still came along — and any cell you cleared, or that came back empty from a spreadsheet round trip, is read as set this metafield to nothing. The values are gone. There is no warning before it happens and no undo after.

It's easy to produce blanks without meaning to. Deleting a row's text to “leave it out”. A filter that hid rows while you retyped a column. A CSV opened and re-saved by a spreadsheet that dropped a value it didn't like. In every case the file looks like the export you started from.

What this tool does instead: blank means leave it alone. A row whose value cell is empty is simply not written into the import file — it can't clear anything, because it isn't there. You still see it: every skipped row is itemised under left unchanged, so “nothing happened” is something you're told, not something you have to infer from an absence. If you genuinely do want to clear a metafield, that's a deliberate delete you make in Shopify — not a side effect of an empty cell.

Footgun 2 — a metafield's type is locked once it holds values

Every metafield has a type: single_line_text_field, number_decimal, boolean, date, and so on. You cannot change that type while the metafield has values. Change the type column in your sheet from a number to text and re-import, and the write doesn't quietly convert — it fails, or the row is skipped, and you're left reading an error that doesn't say “you changed a type”.

The fix Shopify requires is blunt: delete the existing values for that metafield first, change the definition's type, then import the new values. That's a decision worth making on purpose, not one to discover halfway through an import.

What this tool does instead: we read the type on every row and compare it against the type the same metafield uses everywhere else in your file. When two rows disagree about one metafield's type, we hold the odd one back before it's written, name both types and both rows, and tell you the remediation in plain words. Nothing is guessed and nothing is silently converted.

Try it on a file that fires both

Here is a real export-edit-reimport workbook with both footguns loaded: metafields-safe-roundtrip.xlsx. Ten rows: five edited with new values, four deliberately left blank, and one that changes a metafield's type from a number to text. A second sheet records what the store held before the edits, so you can check the result yourself rather than take our word for it.

Run it through the tool above. You should see five rows queued to update, four left unchanged — the blank ones, still holding their old values — and one held back for the type change, with the delete-first remediation spelled out. That is the whole claim of this page, and you can watch it happen on your own screen.

A reviewed file, not an automatic write

This tool never touches your store. It reads your edited file, works out what a re-import would actually do, and hands back a separate import CSV containing only the rows that carry a value you typed. You review it and upload it in your Shopify admin yourself. Your uploaded file is read in memory and never stored, and your original is never modified.

Want this inside your AI agent, on your own machine, with no upload at all? Install xlsx-for-ai and ask Claude or Cursor to “check this metafields file before I re-import it.”