Browser tool · CSV / TSV · Free to use
Product feed checker
Find format, identity, price and variant-group problems in a product CSV or TSV.
Practice datasets are fictional and labeled as such. Real store captures are dated observations with their own sources and evidence limits.
Works in your browser
Check your own data
No signup or upload to a server. Up to 10 MB per file and 50,000 rows or URLs per set. Results describe the supplied files only.
Export a bounded product sample
Supply id, title, link, image_link, price and availability columns. Optional checks use condition, identifier_exists, gtin, brand, mpn, item_group_id, color, size, material, pattern, age_group, gender, availability_date and sale_price. Preserve stable item IDs and use one row per submitted item. Try the sample before uploading a large export.
The checker accepts a plain-title CSV or tab-separated export, not XML feeds, structured_title fields or live integrations. It accepts up to 50,000 rows and 10 MB per file. Quoted commas, quoted newlines and a UTF-8 byte-order mark are supported. Missing columns or inconsistent row widths produce an explicit error.
Know exactly what is checked
Checks cover core-field presence, duplicate IDs, ID/title length, URL and offer formats, supported condition and identifier flags, supplied GTIN checksums and variant combinations. Cross-field checks compare sale prices and identifier evidence. The tool does not validate every attribute in the product data specification.
A syntactically valid currency code is not proof of country eligibility. A valid GTIN checksum is not proof that the identifier belongs to the product. A missing GTIN is not automatically an error for every product. This tool intentionally avoids inventing identifiers or treating every product category as having identical requirements.
Work the sample: two file fixes and one catalog review
The sample’s blue size-nine shoe uses “unavailable.” The checker flags that value because the supported feed enum uses out_of_stock for this situation. Confirm the actual stock state in the commerce platform before making the correction; replacing the word without evidence can turn a format fix into a factual error.
The red size-eight row submits a sale price of 99 USD against a regular price of 89 USD. Confirm the real promotion: correct its mapped amount, or remove sale_price if there is no sale. The red size-nine row repeats Red / 8 despite its title. Compare its source variant with row 4 before correcting size; a title alone is not proof. Keep blue size eight unchanged as the control. These fictional examples illustrate investigation, not live products or approved offers.
Turn the report into an implementation queue
Start with Fix to review supported file-level problems. Review findings identify a conflict or incomplete variant evidence that needs a human decision; they do not prove a disapproval. Filter by field to see whether one mapping explains several items, then search an item ID to inspect its remaining findings. The overview counts affected rows separately from findings so several issues on one item do not look like several products.
The export retains every finding in priority order, even when the screen is filtered. It includes record number, item, field, observed value, action, correction guide and a related record where available. Assign an owner, change Status from Open only after verification, and record the regenerated export or processed-item evidence. Spreadsheet formula prefixes are escaped; this is an issue register, not a corrected feed to import.
Row numbers count non-empty CSV records including the header. A multiline quoted field remains one record. Use both the row and item ID to locate a problem, especially when the finding itself concerns duplicate or missing IDs. Supplied product data stays in the browser; download the issue register before leaving if you need to retain it.
Keep market and category checks separate
The file alone cannot establish the target country's currency, shipping or tax setup, nor which category-specific attributes apply. Confirm the market, language, destination and product category in Merchant Center before reviewing those requirements. Missing optional columns are listed as coverage gaps; that list is not a declaration that every column is mandatory.
A missing GTIN is not automatically a defect, and identifier_exists=false is not a workaround for an unknown identifier. Establish whether the manufacturer assigned identifiers before changing that flag. A supplied flag that conflicts with a GTIN or brand/MPN pair is a review item. Even a valid checksum still needs a product-identity check.
Descriptions, structured titles, shipping, tax, image dimensions and availability-date validity are outside this version's rule set. The date check only detects absence when preorder or backorder is supplied. A zero-finding report therefore means that supported checks found nothing, not that the whole feed is complete or eligible.
Verify the channel after repairing the source
Correct the owning mapping or catalog field, regenerate the file and rerun it. Then use the Merchant Center playbook to inspect processing and landing-page consistency. The checker does not crawl links, test images, inspect account policies or validate every country/category requirement.
Pass when file findings are resolved or documented and the channel-specific checks have separate evidence. A report with no findings is not an approval guarantee. Downloaded findings include row numbers and next actions; the source file is not sent to the server or analytics.
Sources and maintenance
Check current platform documentation before implementation. Review these instructions when your platform, catalog behavior or source guidance changes.