Feed Health Score explained
What the /tools Feed Health Score measures, why feeds above 90 outperform, and exactly which product data problems drag your score down.
Your Feed Health Score is one number, 0–100, that tells you how ready your product data is for Google. It weighs the fields Google cares about most, identifiers, titles, categories, images, pricing accuracy, across every product in your feed. Merchants who get above 90 see measurably more impressions and sales, because Google rewards complete, trustworthy data.
Why one score?
Product feed problems hide in aggregates. A 5,000-product catalog might be "mostly fine" and still have 800 products invisible on Google because of missing GTINs, and another 400 losing every relevant auction because their titles don't contain what shoppers type. The Feed Health Score exists so that "mostly fine" has a number: and so you notice when it moves.
What the score weighs
The score is a weighted composite across every product, driven by the same priorities Google's own systems reward:
- Identifiers: GTIN, MPN and brand present and valid. The single heaviest factor: identifier-complete products get matched to Google's product catalog and win richer placements.
- Titles: length, keyword position, attribute coverage (brand, colour, size, model where relevant), no wasted characters.
- Categorisation: a specific
google_product_category, not a top-level guess. "Apparel & Accessories" scores; "Apparel & Accessories > Clothing > Dresses > Cocktail Dresses" sells. - Descriptions & images: spec-compliant, policy-safe, no placeholder images.
- Attribute completeness: colour, size, gender, age group, material where the category demands them.
- Consistency & risk: price/availability signals in order, variant relationships intact, nothing that pattern-matches known disapproval triggers.
Deep dive How disapproval prediction feeds the score
Beyond static completeness checks, /tools scores each product against patterns learned from real Merchant Center disapprovals: identifier/brand mismatches, category–attribute combinations that trigger manual review, image URLs prone to crawl failures, and title constructions that trip editorial policy. Products flagged as at-risk reduce the score before Google ever disapproves them: the point is to fix the feed pre-emptively rather than react to Diagnostics after impressions are already lost. Predicted-risk items are listed with the specific rule they matched, so you can see exactly why each one was flagged.
Reading your breakdown
Under the headline number, the score breaks down per field and per product. The practical workflow:
- Sort by impact: the app orders fixes by how many products each problem affects and how heavily that field is weighted.
- Run enrichment on the biggest blocks first. Titles and categories usually move the score most in one pass.
- Handle the residue: the products enrichment couldn't confidently fix are listed with a reason each.
Score movement over time
The score is monitored continuously, not just at sync time. Upstream changes (a bulk edit in Shopify, a supplier price file, a platform migration) show up as score movement with a change log entry, so a quiet data regression can't cost you weeks of performance before anyone notices.
Frequently asked questions
What's a good Feed Health Score?
Above 90 is the target. Feeds above 90 have seen lifts of over 100% in attributable sales versus their pre-enrichment baseline. Between 70 and 90 means Google is showing you, but you're losing auctions you could win. Below 70 usually means structural problems, missing identifiers or categories, that cap your visibility.
Why did my score drop suddenly?
Usually a change upstream, a product import that wiped barcodes, a price sync issue, or new products added without complete data. /tools monitors your feed continuously and the change log will show exactly what moved and when.
Does enrichment fix my score automatically?
Enrichment fixes what can be fixed from your data and trusted external sources, titles, categories, normalised attributes, missing GTINs where they can be confidently matched. A small remainder (e.g. genuinely unlabelled products with no barcode anywhere) may need your input, and the app tells you exactly which products those are.