Before you upload a product feed: check the row, the variant and the page

A product file can look tidy and still describe the wrong offer. The useful question is whether each row points to the product a customer can actually buy, at the price and availability you submitted.

Start with a copy of your catalog export. Keep the original and record when it was exported. The following is a short review workflow for English product data and US-dollar offers. It is a starting point, not a complete Merchant Center compliance checklist.

First confirm which file you are checking

A current-looking spreadsheet can still cover only a selected set of products, or differ from the file Merchant Center last received. Keep the export time and the source version with your notes.

• Working from Shopify: use the export selection and coverage checklist to distinguish the rows you selected from the full catalogue. A Shopify export is source evidence, not automatically a ready-to-upload Google feed.
• The file looks right but an issue remains: check the last file Merchant Center actually received before treating your latest local copy as the submitted version.

These are free, self-directed checks. File receipt, successful parsing and product approval are different outcomes.

1. Keep the product identity intact

Sort your file by id and check for duplicates and blank IDs. Preserve stable IDs when correcting an existing item: a fresh export should not make yesterday's product appear to be a different product. Google advises using a unique, stable identifier for each product. Google: ID

Then check the relationship between the ID and the variant. A blue mug and a green mug may share a product page, but a row for the blue version needs to describe that version consistently.

2. Separate formatting from facts

This deliberately incomplete CSV excerpt is illustrative. The products, prices and URLs are invented teaching examples, not customer data, a production feed or measured results.

id,title,price,availability,link
MUG-BLUE,Blue ceramic mug,18.00 usd,In Stock,https://example.com/products/mug?color=blue
MUG-GREEN,Green ceramic mug,18.00 USD,in_stock,https://example.com/products/mug?color=green

For the first row, changing 18.00 usd to 18.00 USD and In Stock to in_stock standardizes the format while preserving the stated meaning. Google's specifications use an amount with a currency code and supported availability values such as in_stock. Google: price, Google: availability

That edit does not prove the mug costs USD 18 or is available. If the store actually shows USD 20, check which system is authoritative and update the incorrect source. Do not pick a price because it makes the file pass a check.

3. Open the exact variant URL

Open the submitted link in a fresh browser session. Check the selected variant, visible price, currency and availability. Google's link guidance recommends selecting the correct variant on arrival. Google: link

Record what you saw and when. Compare the same variant across the file, page and checkout. Google requires matching price information and matching availability across the relevant surfaces. A promotion, stale export or default variant can make a seemingly correct row misleading. Google: price, Google: availability

A failed page fetch is a separate outcome. Mark it for another check; do not turn “could not inspect” into “the merchant has no price.”

4. Leave unknown identifiers unresolved

Do not fill an empty GTIN cell with a SKU, a random barcode or another variant's identifier. Check reliable manufacturer or supplier records. Google explicitly warns against guessing GTINs; some products genuinely have no assigned GTIN. Google: GTIN

Use an internal “needs source confirmation” list. Keep that workflow label separate from the product value you eventually submit.

5. Make every change reviewable

Keep a small change log: product ID, field, old value, new value, reason and source. Distinguish a formatting correction from a factual correction. Recheck the edited rows, keep the original export, and test your import process before applying a wider batch.

Also record what you did not check. Inspecting five pages does not prove that all fifty product pages match their rows. A useful report states its coverage and outstanding questions.

Mingche Commerce provides automated product-data audits for up to 200 supplied records and supported CSV repairs. Start with our free three-page check, or compare the audit and repair options. Current paid scope: Google, the US market, English data and USD offers. We deliver findings and files for you to use; platform approval remains Google's decision.

Prepared with AI assistance using the official sources linked above. Examples are illustrative.