Most ecommerce teams set up their product feeds once, then don't look at them again until something breaks. A quarterly feed audit catches issues that silently erode performance — wrong prices, missing GTINs, thin descriptions, outdated categories — before they cost you significant revenue.
This checklist covers Google Shopping, Meta, and ChatGPT feeds. Work through each section methodically — it typically takes 2-3 hours to complete a thorough audit of a mid-size catalogue.
Section 1: Data completeness
Start with the basics — are all the fields that should be populated actually populated?
- ✅ Every product has a title (non-empty)
- ✅ Every product has a description (minimum 100 characters recommended)
- ✅ Every branded product has a brand field
- ✅ Every product has at least one image URL
- ✅ Every product has a price
- ✅ Every product has an availability status
- ✅ Every product that has a barcode has it mapped to gtin
- ✅ Products without barcodes have identifier_exists: no set
- ✅ Apparel products have gender and age_group
- ✅ Variable products have item_group_id (or group_id for ChatGPT)
Section 2: Data accuracy
Correct data is as important as complete data.
- ✅ Feed prices match website prices (spot-check 10 random products)
- ✅ Availability statuses match website stock (check 5 in-stock and 5 out-of-stock)
- ✅ All product URLs return HTTP 200 (not 301 or 404)
- ✅ All image URLs return accessible images (not 404)
- ✅ Images meet minimum size requirements (250×250px for Google, 500×500px for Meta)
- ✅ GTINs pass checksum validation (13-digit EAN format for UK)
- ✅ No duplicate item IDs in the feed
- ✅ Sale prices are actually lower than regular prices
Section 3: Title quality
- ✅ Titles include brand name
- ✅ Titles include key variant attributes (colour, size, material where relevant)
- ✅ No marketing language in titles ("Amazing", "Best Quality", "Sale")
- ✅ Titles are at least 40 characters (underutilised titles miss match opportunities)
- ✅ Titles don't exceed 150 characters
- ✅ Most important attributes appear in the first 70 characters
Section 4: Description quality
- ✅ No HTML markup in descriptions (Google requires clean text; ChatGPT will reject HTML)
- ✅ No non-ASCII characters (curly quotes, em dashes, copyright symbols)
- ✅ Descriptions are factual and specific (not just marketing copy)
- ✅ Descriptions mention key product specifications (dimensions, materials, compatibility)
- ✅ No duplicated descriptions across multiple products
Section 5: Category mapping
- ✅ google_product_category is populated for all products
- ✅ Categories are specific (level 4+ in Google's taxonomy, not just level 1)
- ✅ Product_type field is populated with your own category hierarchy
- ✅ No products sitting in catch-all categories like "Other"
Section 6: Custom labels
- ✅ Custom label fields are being used (at least one of custom_label_0-4)
- ✅ Labels are being used actively in campaign bid management
- ✅ Label values are consistent (not "high-margin" and "High Margin" mixed)
- ✅ Labels reflect current product status (clearance products labelled, seasonal updated)
Section 7: Feed infrastructure
- ✅ Feed is set up as a scheduled fetch URL (not a manual upload)
- ✅ Feed updates at least daily
- ✅ Last successful run was within the last 24 hours
- ✅ Email alerts are configured for failed feed runs
- ✅ Feed has not been manually edited in a spreadsheet recently (manual edits break rules)
Section 8: ChatGPT-specific checks
If you're running or building a ChatGPT feed:
- ✅ is_eligible_search is set to true for all appropriate products
- ✅ variant_dict only contains dimensions that genuinely vary per product group
- ✅ group_id correctly groups all variants of the same product
- ✅ All descriptions are free of HTML and special characters
- ✅ Price format is "12.50 GBP" (number space currency code)
- ✅ seller_url and return_policy are populated and accurate
Automate your feed checks
Trajekt validates feeds against common issues before every submission — no manual checklist needed.
Start free →What to do with audit findings
After completing the audit, prioritise fixes by impact:
- Critical (fix immediately): Price mismatches, missing required fields, broken URLs, HTML in ChatGPT descriptions
- High priority (fix this week): Missing GTINs, thin descriptions, missing variant grouping
- Improvement (fix this month): Title optimisation, category refinement, custom label updates
Most fixes can be applied via feed rules without touching your website — rule-based title reformatting, field concatenation, value substitution. For data issues that originate in your ecommerce platform (wrong prices, missing barcodes), the fix needs to happen at source.
Run this audit quarterly and you'll maintain feed quality as a sustainable competitive advantage.