There’s a moment every performance marketer faces the same paradox: budget increased, bids optimized, audiences well-built — and yet campaigns won’t scale. ROAS stays flat, PMax concentrates spend on five products while ignoring the other nine hundred, Shopping CTR is low for no apparent reason.
In most cases, there is a reason. And it’s in the feed.
The product feed is the primary signal Google and Meta use to make targeting, bidding, and creative decisions in algorithmic campaigns. Before touching bids or revisiting audiences, it’s always worth asking: is the feed truly optimized to perform, or is it just technically correct?
This article answers that question with concrete, actionable guidance.
Why optimizing the feed is different from optimizing a campaign
Classic campaign optimization works on levers external to the product: bids, budgets, audience segments, creatives. Product feed optimization works on the product data itself — and has an impact that cascades across every campaign that uses that data as input.
An optimized product title improves relevance in Google Shopping, boosts signal quality for PMax, and increases relevance in Meta’s Dynamic Product Ads. Conversely, a non-compliant image gets disapproved in Merchant Center and blocks the campaign upstream. A missing GTIN reduces eligibility for free Shopping ads across the entire catalog.
The feed is not a campaign setting: it’s the infrastructure campaigns run on. Optimizing it has multiplying effects that no bidding lever can replicate.
The five attributes with the greatest impact on performance
Not all feed attributes carry the same weight. These five are the ones to tackle first, in order of impact.
1. The product title
The title is the attribute with the greatest direct impact on query relevance and algorithmic relevance. Google uses it as the primary signal to match products to user queries in Shopping and PMax.
An optimized title follows a precise structure that varies by product category, but the general principle is the same: the most important information must appear in the first 70 characters, because that’s what’s shown in the ad before truncation.
For apparel, the optimal structure is: "Brand + Product type + Key attributes (color, material, size)" Levi’s Jeans Slim Fit Uomo Blu Denim 511
For electronics: "Brand + Model + Main technical specs" Example: Samsung Galaxy S24 Ultra 256GB Titanium Black
What to avoid: promotional terms (“Special offer”, “Best seller”), special characters, excessive capitalization, redundant information already present in other attributes.
2. The GTIN (Global Trade Item Number)
The GTIN — EAN, UPC, or ISBN code — is the product’s universal identifier. Google uses it to connect your product to its global catalog database (Shopping Graph), which contains aggregated information on prices, reviews, and availability from millions of sources.
A product with a correct GTIN automatically gains greater visibility: it accesses free Shopping ads, is more easily matched to purchase queries, and receives a relevance boost in PMax campaigns. A product without a GTIN is treated as unverified data — resulting in reduced ad eligibility and lower auction priority.
For brands that manufacture their own products without a standard GTIN, Google accepts the "identifier_exists = false attribute" — but it must be declared explicitly, not omitted.
3. Custom labels
Custom labels are five free-form attributes (custom_label_0 through custom_label_4) that don’t affect search relevance, but are the most powerful tool for segmenting the catalog within Shopping and PMax campaigns.
Each custom label can contain any value you choose: product margin (high/medium/low), seasonality (spring/summer/fall/winter), historical performance (top seller/slow mover/new arrival), price range, or any other dimension relevant to your strategy.
In PMax, custom labels allow you to create separate asset groups for different catalog segments, applying differentiated audience signals and creatives. Without custom labels, PMax treats the entire catalog as a single block and concentrates spend where the algorithm finds the strongest existing signals — typically products that already convert, leaving a significant portion of the catalog unexplored.
4. Images
Images in the feed serve two distinct functions: meeting technical requirements for approval, and maximizing CTR where they appear.
For approval on Google Merchant Center, the minimum requirements are: white or neutral background for apparel, no overlaid watermarks or text, minimum size 100x100 px (recommended 800x800 px or higher for high-resolution display).
For CTR on Meta and TikTok — where the image is the first attention filter — the rules change: lifestyle images (worn, in context) outperform images on a neutral background. This creates tension with Google’s requirements, which call for a neutral background. The solution is to have two image variants in the feed — one for Google and one for Meta — managed by the feed management system.
One often-overlooked aspect: Google prefers 1:1 ratio images for maximum coverage across all placements. Vertical (4:5) or horizontal images are accepted but receive reduced coverage on some ad formats.
5. Real-time availability and price
Availability and price are the attributes with the most immediate impact on disapprovals and on the quality of data as perceived by the algorithm.
A product flagged as available in the feed but out of stock on the site triggers an automatic disapproval on Merchant Center within 24–48 hours. The more disapprovals accumulate, the worse the account’s trust score becomes — with negative effects on the entire catalog, not just the products involved.
A feed price that doesn’t match the landing page price triggers the same disapproval. During promotions or sales, this happens frequently if the feed isn’t updated in real time or with sufficient frequency.
The practical rule: for catalogs with frequent price or availability changes, updating the feed at least twice a day is the minimum. For high-turnover stock, real-time updates via API are necessary.
How to optimize the feed specifically for Performance Max
PMax deserves a separate deep dive because its relationship with the feed differs from classic Shopping.
In Shopping, the feed primarily serves to match products to queries. In PMax, the feed is the main input on which the algorithm builds its decisions about who to target, which creative to show, and which placement to invest budget in. There are no negative keywords, no mandatory audience segments: the feed is what speaks to the algorithm.
Three PMax-specific practices:
Segment the catalog with asset groups. Creating separate asset groups by product category, price range, or seasonality — using custom labels as the segmentation basis — allows you to associate specific creatives and audience signals with each segment. A single asset group for the entire catalog is the configuration that produces the worst results.
Exclude unprofitable products. PMax tends to spend across the entire catalog indiscriminately during the learning phase. Excluding low-margin products, out-of-stock products, or products that historically don’t convert reduces budget waste in the early phase and helps the algorithm converge faster on profitable products.
Enrich text attributes for low-coverage categories. Products with the most incomplete feed data receive less PMax budget not because the algorithm explicitly penalizes them, but because it has fewer signals to work with. Completing titles, descriptions, and category attributes for these products is often the intervention with the best impact-to-effort ratio.
How to measure feed quality
Before optimizing, you need to know where the feed’s most serious problems are. Three tools to use in combination:
Google Merchant Center — Product diagnostics. Shows all disapproved products, disapproval reasons, and products with warnings (not disapproved, but with missing or suboptimal attributes). This is the mandatory starting point: resolve approval issues before working on attribute optimization.
Google Merchant Center — Competitive visibility. Shows benchmarks for Click Share and Impression Share relative to competitors in the same product category. If your Click Share is low despite competitive bids, the problem is feed quality, not bidding.
GA4 — Performance by item_id. With eCommerce tracking properly configured, GA4 lets you analyze conversions by individual product, cross-referenced with feed data. This is the most direct way to identify products receiving impressions but not converting — often a sign of an irrelevant title or image.
The three-phase optimization process
Phase 1 — Audit and error correction (highest priority) Resolve all active disapprovals on Merchant Center. While products are disapproved, the account accumulates penalties that spread across the rest of the catalog.
Phase 2 — High-impact attribute optimization (high priority) Work on titles, GTINs, and images for products with the highest commercial potential — typically those with high margins or high search volume in their category.
Phase 3 — Segmentation and custom labels (medium priority, structural effect) Build the custom label structure that allows strategic catalog management across PMax and Shopping campaigns. This is the intervention with the most lasting impact, but it requires the most planning.
Conclusion
Product feed optimization is not a one-time activity: it’s an ongoing process that requires monitoring, updates, and regular interventions as the catalog changes and platforms evolve their specifications.
The starting point is always the same: understand where the feed is causing problems before looking for solutions in campaign settings. An optimized feed doesn’t eliminate the need for a solid bidding strategy — but it creates the conditions for that strategy to work.
→ Discover Product Feed Marketing suite by Highstreet.io
→ Performance Feed & Analytics
→ Request a demo