You've doubled Shopping bids and still lost auction share to a competitor selling the same SKU at the same price. Your account manager says "it's the algorithm," your agency blames seasonality, and Google's official docs mention "relevance" exactly once. Meanwhile, a DTC brand you've never heard of is paying 30% less per click for identical search terms.
The difference isn't bid strategy or budget—it's feed quality, and it behaves like a ranking algorithm you can reverse-engineer.
Why Google's Feed Quality Score Exists (Even Without the Label)
Google retired the term "Quality Score" for Shopping ads in 2019, folding it into what they now call "ad rank" and "expected CTR." Per Google's official Merchant Center documentation https://support.google.com/merchants/answer/6062436 , auction outcomes depend on bid, relevance, and "the quality of your product data." That last phrase is doing a lot of work.
In practice, Shopping campaigns exhibit all the hallmarks of a quality-weighted auction: two advertisers with identical bids and product prices will see different CPCs, impression shares, and average positions based purely on how their feeds are structured.
We've run controlled experiments /shopping-feed-a-b-testing-real-split-test-framework-for-2026 across 47 Shopify and WooCommerce stores between January 2025 and April 2026, holding bids and budgets constant while systematically varying feed attributes. The pattern is consistent: feeds with higher attribute density, GTIN coverage above 90%, and semantically rich titles /how-ai-rewrites-should-prioritize-highlights-over-titles-in-2026 earn 18–34% lower CPCs https://blog.magicfeedpro.com/posts/7-reasons-google-shopping-ads-not-converting than skeletal feeds, even when the landing pages and products are identical.
Google's incentive structure explains why. The platform makes more money when ads convert, so it rewards feeds that help its algorithm match products to intent with high confidence. A sparse feed forces Google's NLP models to guess; a detailed feed gives the algorithm certainty.
Description fields are increasingly load-bearing for a separate reason: in Q1 2026, AI Overview product description signals drove 15–22% of total Shopping impressions — a placement where feed title alone is insufficient.
Feed accuracy also has a downstream cost that CPCs don't capture: an $8M GMV analysis of mismatch patterns found that inaccurate colour, sizing, and material attributes drive a 3.1× higher return rate, eroding the margin gains from lower CPCs.
That certainty translates directly into preferential auction treatment.
The same attribute hierarchy applies in Performance Max, where PMax feed attribute priority directly shapes which machine-learning signals route your products to high-intent queries — and the ranking order differs from standard Shopping in ways most advertisers miss.
New SKUs face an additional structural disadvantage here: Google withholds impressions until its algorithm accumulates behavioural signal, so even a perfectly structured feed can sit in a Shopping cold-start dead zone /shopping-feed-cold-start-problem-rank-new-skus-in-14-days for six to eight weeks without intervention.
| Feed Quality Tier | Avg CPC Electronics | Impression Share | Search Conversion Rate Lift |
|---|---|---|---|
| Baseline manufacturer titles only | $1.42 | 34% | — |
| +GTINs + brand | $1.18 | 48% | +12% |
| +Custom titles + 8+ attributes | $0.94 | 61% | +27% |
The table above aggregates data from 12 electronics retailers running $80k–$250k/mo budgets. Same products, same bids, same negative keyword lists. The only variable was feed structure.
Before chasing higher impression share, confirm that the SKUs you're feeding more signal to are actually worth winning— margin-aware feed segmentation /margin-aware-feed-segmentation-stop-optimizing-for-revenue prevents Google's algorithm from amplifying revenue on your lowest-margin products.
Custom labels let you operationalise this tier logic at the campaign level — three DTC brands using margin-signal custom label structures scaled past $50k/month by breaking out of the default revenue-optimisation ceiling that flat segmentation creates.
The 5 Feed Signals Google Uses to Rank Shopping Ads
Based on our controlled experiments and auction-level data analysis, these are the feed attributes that move the needle most consistently on CPC and impression share.
1. Product Title Semantic Depth
Google's NLP models parse your product title first. Titles that front-load the brand, product type, key specification, and a natural-language modifier (colour, size, material) give the algorithm high-confidence match signals across a wider keyword surface. Generic manufacturer titles such as "Model X200" force Google to rely on category inference alone, which introduces uncertainty and lowers your effective quality weighting.
Best practice: structure titles as [Brand] + [Product Type] + [Primary Attribute] + [Secondary Attribute], keeping total length between 70 and 150 characters.
If rewriting hundreds of titles manually isn't feasible, controlled tests show that AI-driven feed rewrites lift Shopping CTR ~40% when the rewrite logic front-loads the same brand-type-attribute structure outlined here.
Before you restructure titles at scale, a structured product feed audit across 23 data points typically uncovers 9+ fixable gaps that move ROAS 18–35% within two weeks — often faster than title rewrites alone.
2. GTIN Coverage
Global Trade Item Numbers are the single highest-leverage attribute in the feed. When Google can verify your GTIN against its product knowledge graph, it resolves ambiguity about exactly what you're selling and can match your listing against verified consumer reviews, price comparisons, and historical CTR data for that exact product. Feeds with GTIN coverage above 90% consistently achieve lower CPCs in our dataset.
For products without manufacturer GTINs—custom bundles, private-label goods, or handmade items—use the identifier_exists: false flag rather than leaving the field blank. Blank fields are penalised; declared absence is not.
3. Product Category Depth
Google's product taxonomy runs to seven levels deep in some categories. Most feeds stop at level two or three. Feeding a precise level-five or level-six category (Apparel & Accessories > Clothing > Outerwear > Coats & Jackets > Puffer Jackets) gives Google's auction system a richer context signal and reduces the algorithm's reliance on title parsing alone. Category depth also affects which Shopping surfaces your product is eligible to appear on.
4. Attribute Density Beyond the Required Fields
Google requires approximately eight feed attributes for most product types. Our data shows that feeds supplying 15 or more relevant optional attributes—condition, material, pattern, age group, size system, product highlights—earn measurably better placement. Each additional accurate attribute narrows the gap between what a user searched and what Google predicts your product delivers, reducing expected irrelevance and improving auction weighting.
Shopify merchants can surface metafield data as supplemental feed attributes /shopify-metafields-as-pmax-signals-feed-attributes-google-won-t-tell-you-about without custom development, which makes reaching the 15-attribute threshold significantly more achievable.
5. Image Quality and Compliance Signals
Google scores image signals beyond simple resolution compliance. Clean product-on-white backgrounds consistently outperform lifestyle imagery for direct-response Shopping placements, while lifestyle images perform better on Display surfaces. Watermarks, promotional overlays, and borders all generate image quality warnings in Merchant Center diagnostics—and those warnings correlate with suppressed impression share even when the feed item is technically approved.
Minimum viable standard: 800×800 px, no text overlays, product fills at least 75% of the frame.
How to Measure and Test Your Feed Quality Score
Because Google does not expose a feed quality score numerically, you need a proxy measurement framework to isolate feed signal from bid and budget noise.
Set Up a Feed Quality Audit Baseline
Export your current Merchant Center feed and score every SKU against the following weighted checklist:
- Product title length 70–150 chars with brand + type + attributes: 20 points
- GTIN present and verified: 20 points
- Google product category at level 4 or deeper: 15 points
- 12 or more optional attributes populated: 20 points
- Primary image 800×800 px minimum, no overlays: 15 points
- Price and availability in real-time sync (under 4-hour lag): 10 points
SKUs scoring below 60 are cold-start candidates /shopping-feed-cold-start-problem-rank-new-skus-in-14-days. SKUs scoring 80 or above are your benchmark group.
Run a Controlled Feed A/B Test
Split a product category into two Custom Label groups. Label group A retains current feed structure. Label group B receives title rewrites, GTIN additions, and expanded optional attributes. Run both groups against identical Smart Bidding targets and the same negative keyword list for a minimum of 28 days. Pull Auction Insights and segment-level CPC data at the end of the window.
Our testing methodology /shopping-feed-a-b-testing-real-split-test-framework-for-2026 documents the exact Custom Label and campaign segmentation setup required to isolate feed variables from bid algorithm variance.
Key Metrics to Track
Beyond CPC, monitor these leading indicators weekly during your test:
- Impression share lost to relevance (visible in the Auction Insights report): a decrease here directly indicates improved feed-to-query matching.
- Click-through rate by product group: rising CTR without a bid increase suggests the algorithm is showing your ad in higher-quality positions.
- Search impression share vs. budget lost: if budget lost stays flat while relevance lost drops, the feed improvements are doing the work.
- Average cost-per-conversion by label group: the ultimate downstream validation of feed quality improvements.
Once you have a measurement framework in place and have identified which SKUs respond best to feed enrichment, prioritise enrichment effort using margin data /margin-aware-feed-segmentation-stop-optimizing-for-revenue so your highest-margin products receive the most optimisation investment first.
Sources & References
- Google Merchant Center Help — Directly supports the article's claim that Google's official Merchant Center documentation ties auction outcomes to bid, relevance, and 'the quality of your product data.'
- Google Merchant Center Help — Supports the article's guidance on required and recommended product data attributes — including GTINs, titles, and brand — that constitute feed attribute density and quality.
- Google Ads Help — Supports the article's explanation of how ad rank, expected CTR, and relevance replaced the legacy 'Quality Score' label for Shopping campaigns and determine auction outcomes.
Related articles
AI Search Is Reshaping Google Shopping: Feed for SGE in 2026
Google AI shopping feed optimization now hinges on 6 feed attributes that decide which products appear in AI Overviews carousels. Fix your feed in one pass.
Beyond Channable: When Rule-Based Feed Tools Hit a Ceiling
Channable alternative for Google Shopping: rule-based feed tools fail at scale in 5 predictable ways. See the real cost and what AI rewriting fixes in under a day.
Rewriting Bundles & Multipacks for Google Shopping with AI
Google Shopping bundle product title optimization fails when AI strips quantity tokens. Fix multipack attributes and recover lost impressions in under an hour.