How feed quality and AI product enrichment control your margin
Anyone looking to optimize Google Shopping almost always starts with their bids. Wrong place. Your shopping ads are displayed based on the data in your product feed – titles, attributes, categories, and prices. If this feed is sparse, even the best bid won’t generate profitable clicks. This article shows where optimization really begins, the five feed weaknesses that hold most shops back, and how we use the Feed Creator and its AI-powered product enrichment to get your feeds in shape.
Where Google Shopping optimization really begins – with the feed, not the bid.
The visibility of your Shopping ads depends on the data quality of the feed before the first bid is applied. Google compares every search query against your product data. What’s not in the feed can’t be found – and what’s poorly positioned in the feed will be displayed at an inflated price.
Why bid optimization is wasted on a weak feed
Bidding strategies like Smart Bidding learn from the data they receive. If you provide them with generic titles and missing attributes, the algorithm learns from a poor foundation. You then pay for clicks that rarely convert. Only a clean feed gives the bid management the data it needs to protect your margin.
The three key levers: title, attributes, structure
Three factors determine the majority of feed quality. The title determines for which search queries you even appear. The attributes (brand, GTIN, size, color, material) determine how precisely Google categorizes your product. The structure – product types and categories – controls how you can allocate campaigns and budgets. You’ll gain the most leverage by focusing on these three aspects, not the daily bid. What exactly is included under “attributes” – from the GTIN and variants to product highlights and product details – is explained in the next section.
The 5 most common feed weaknesses that slow down your shopping performance
Five weaknesses recur in almost every audit. They can all be measured and corrected.
Thin or generic product titles
Titles like “women’s blue sweater” miss out on reach. They lack brand, material, cut, and occasion – precisely the terms that customers ready to buy are searching for. Those who refine their search terms will appear in more and more relevant search results.
Missing GTIN/brand and incorrect categories
Without a GTIN and brand, Google ranks products lower and displays them less frequently. Incorrect Google product categories mean you’re competing against the wrong competitors. Both of these things reduce visibility, without being immediately noticeable in your account.
Empty or copied descriptions
Standard descriptions taken directly from the shop or empty fields provide Google with little context. The description is a ranking signal – filling it out is worthwhile, especially for large product ranges where no one manually updates every description.
Outdated prices and availability
If the price in the feed doesn’t match the price on the landing page, the Merchant Center flags the product – in the worst case, it’s removed. Outdated availability wastes budget on items that aren’t even available.
No sell signal in the feed (all products treated equally)
Most feeds treat the bestseller like the slow-moving item. Performance Max then allocates the budget according to its own discretion, not based on your profit margin. There’s no signal indicating: this product deserves more budget, that one less.
Which feed attributes really matter – from mandatory features to product highlights
Most online shops fill in the required fields and then stop there. That’s precisely where they’re missing out on valuable reach. Google compares every search query against your product data – and the more complete and relevant this data is, the more likely a product is to appear for the relevant searches. Reach and profit margin are determined by these attributes, not by the daily bid.
The title – your most important line
The product title can be up to 150 characters long, but depending on the device, usually only the first 70 or so are visible. Therefore, the most important information should come first: brand and product type first, followed by the distinguishing features that customers ready to buy are looking for. What these features are depends on the category – for clothing, size and color; for electronics, storage capacity and color; and for subscriptions, the contract duration. What doesn’t belong in the title: all caps for emphasis, advertising phrases like “Sale” or “Free Shipping,” and symbols – Google even interprets all caps as a spam signal.
The description – substance over advertising language
The description can be up to 5,000 characters long, but the beginning is crucial: the most important features should be in the first 160 to 500 characters, before the reader has to expand the text. Write about what makes the product special – material, dimensions, suitability, variations – not how great your shop is. Prices, shipping information, or the company name have no place in the description.
Identifiers – GTIN, MPN and brand
gtin,mpn andbrand Tell Google what you sell. If the GTIN is missing, Google explicitly limits visibility – a clean assignment “cannot be guaranteed”. For new branded products, the GTIN is practically mandatory. Where one doesn’t exist, the MPN steps in; the brand is mandatory for new products anyway. The fieldidentifier_exists Setting it to “no” is only correct in truly exceptional cases – unique pieces, handmade items, vintage goods. For regular branded goods, it’s an expensive mistake.
Correctly represent structure and variants
Two fields categorize your product: thegoogle_product_category (Google’s own taxonomy) and the freely selectableproduct_type for your bidding and reporting logic. Variants come into play.item_group_id In addition, it groups color, size, or storage variations for a single product and is mandatory for product variants in Germany. For clothing, Google requires an additional [feature/component] in Germany.color ,size ,gender andage_group These attributes feed the filters in the search results – if they are missing, your product won’t even appear in the relevant facets.
Product Highlights – the underestimated selling points
The attributeproduct_highlight It’s a short list of the strongest buying arguments – and in most feeds, simply empty. Google recommends four to six highlights per product, each up to 150 characters. Important: These should highlight advantages, not just technical data and avoid repeating the title or description. Google is increasingly displaying these highlights on AI-powered interfaces and in the search tab – a feature that hardly anyone is currently using.
Product Detail – structured technical data
Where Product Highlights convey the benefits, it deliversproduct_detail The hard facts: structured technical specifications in the formAbschnitt : Merkmal : Wert – for exampleKonnektivität : WLAN-Standard : 802.11b/g/n Each attribute can contain up to 1,000 characters. This field is intended for precisely those details that no other attribute covers and feeds the same AI and search interfaces as the highlights. Especially for products requiring explanation—technology, spare parts, features—this separates easily discoverable items from those that are invisible.
Images – clean, large, without distractions
The main image must clearly show the product: no watermarks, no logos, no advertising text, no price distractions. From January 31, 2027, Google requires a minimum size of 500 × 500 pixels for all categories – those who submit larger images now are on the safe side and also qualify for placements like YouTube and Connected TV in Performance Max.additional_image_link Up to ten additional images can be added, showing the product from different angles or in use – this increases click relevance.
In what order the work is worthwhile
Feed optimization in practice: the Feed Creator step by step
We optimize shopping feeds with the Feed Creator – an application that moves your product data from source to Merchant Center in five steps. Each step is transparent and addresses one of the five weaknesses.
Step 1: Import – Product data from any source
The Feed Creator reads your product data directly from where it resides: via XML or CSV URL, the Shopify GraphQL API, or the Shopware 6 Admin API. A second source can be added via a merge, for example, to enrich stock levels or margin data. This way, we work with your actual data, not a copy from yesterday.
Step 2: Rules – Normalize titles, descriptions, categories, and GTINs
Over 30 actions are available to normalize, filter, and exclude fields—text, numbers, lists, aggregations, combined via nested AND/OR/NOT conditions. This allows us to clean up titles, standardize categories, add GTINs and brands, and exclude products that shouldn’t be advertised. Rules are applied consistently and reproducibly every time, not manually.
Step 3: AI product enrichment – better titles, descriptions, and attributes
AI product enrichment steps in where data is missing or weak. It uses AI to generate better titles, more complete descriptions, and cleaner attributes – from your existing product data, and only for those products that need improvement. More on this in the next section.
Step 4: Performance Label Engine – giving each product a sales signal
The Performance Label Engine evaluates each product and assigns a label. This allows Performance Max to determine which products deserve more budget and which deserve less. For detailed information on how the scoring works, please see the page where we explain Performance Labels and their control by PMax .
Step 5: Google Upload – cleanly into the Merchant Center
Finally, the Feed Creator uploads the finished data to the Merchant Center – via the Merchant API (the Content API for Shopping will be replaced in August 2026), with SKU diff (only changes are transferred), URL fetch, and optionally via SFTP. This keeps the upload fast and transparent.
What specific improvements AI product enrichment makes to your product data
AI product enrichment uses your existing product data to generate better titles, descriptions, and attributes – AI-powered, only for the products that need it. It doesn’t invent attributes, but rather transforms the data you have into what Google and your customers need.
What changes measurably
“Blue sweater” becomes a title that highlights the brand, material, and cut. Empty description fields are filled, missing attributes are added, and inconsistent spellings are standardized. In a catalog with several thousand items, this is the difference between “theoretically optimizable” and “actually optimized.” Where fields like Product Highlights or Product Details are missing, the enrichment process fills them in as well.
Maintain control: only modified products, review before uploading
The enrichment process only runs on the products that actually need it – this keeps costs down and prevents good content from being overwritten. You can see what will change before uploading. Your account and data belong to you, even while we manage the feed.
From feed to PMax control: Performance labels instead of a scattershot approach.
An optimized feed alone doesn’t guarantee smart budget allocation – each product needs a performance label. The Feed Creator sorts your products into six tiers: HERO (top performers with high margin and ROAS), SOLID (stable, reliable performers), CHANCE (little data, but high potential), SLEEPER (weak current performance, potential exists), COSTLY (margin or POAS below the threshold), and NEW (new, no data yet). These labels are then used ascustom_label_0–4 in the feed and control how Performance Max allocates your budget – according to margin, not with a scattershot approach. How six signal blocks become one powerful tool is explained in our guide on how a clean feed makes Performance Max controllable . How to adjust these values step by step incustom_label_0–4 Our guide to Google Shopping Custom Labels shows you how to write and then split PMax.
What feed optimization brings in practice
Clean feeds and targeted budget management translate into hard numbers – two examples from our work.
A premium water filter shop saw a 215.6% increase in revenue , a 183.3% increase in conversions, and a rise in ROAS from 3.48 to 5.08. A motorcycle parts retailer experienced a 435% increase in revenue , a 235% increase in conversions, and a rise in ROAS from 3.79 to 4.72 – from 1,052 to 3,529 conversions per month. In both cases, the feed was the key driver, not the bid.
Optimize yourself or have it optimized for you?
The mechanics can be replicated – the effort lies in the ongoing maintenance, not the initial setup. Manually improving a single title isn’t rocket science. However, keeping thousands of titles clean, maintaining rules, recalculating performance labels nightly, and ensuring the Merchant Center is error-free is a continuous task. If you lack the time for this maintenance, we’ll take care of it for you. You’ll speak directly with a senior manager, not a junior, and we’ll manage your margin, not just your revenue. We describe the setup and ongoing maintenance process on our page about how we build and maintain your product feed .
Frequently asked questions about Google Shopping optimization
Request free feed analysis
We’ll analyze your shopping feed and show you the biggest levers – no obligation.
