Why Your Product Feed Is Killing Your Shopping Ad ROAS
When Shopping ROAS drops, the first instinct is to adjust bids. The second instinct is to refresh creative. The third, if things are still not moving, is to restructure campaigns. Most teams cycle through all three before anyone looks at the product feed, which is where the actual problem usually lives.
This is a sequencing error. Bid adjustments and campaign structure changes operate on top of the feed. If the feed is providing poor data, those adjustments are optimising a broken foundation. It is the equivalent of tuning the engine on a car with flat tyres. The mechanical work is real, but the vehicle is not going anywhere faster.
An apparel brand managing 12,000 SKUs spent three months adjusting bids on underperforming product groups before a feed audit revealed that 34% of their active SKUs had titles structured as stock codes rather than descriptive product names. Google was matching those products to near-zero relevant queries. The bid adjustments had been optimising impression share on searches that would never convert. Two weeks after restructuring the titles, ROAS on the affected product group increased 41%. The bids had not changed.
The feed is not a technical afterthought. It is the primary signal Google uses to determine which searches your products appear for, how prominently they are ranked within the auction, and how relevant the algorithm considers your listings to be. Every bidding and targeting decision you make operates downstream of that signal.
What Google Actually Does With Your Product Feed
The relevance scoring draws primarily on four feed attributes: the product title, the product description, the product type and category taxonomy, and the GTIN or unique product identifier. Of these, the title carries the most weight by a significant margin. Google extracts keywords from the title to match products to search queries in the same way a search algorithm processes a webpage. A title that is vague, incorrectly structured, or missing key attributes is a title that matches fewer relevant queries, regardless of how competitive the bid is.
The feed is also the source from which Google populates the product listing ad itself: the image, the price, the merchant name, and the title text that appears in the Shopping unit. Poor feed quality therefore creates a dual problem. It reduces the number of relevant queries your products appear for, and it reduces the click-through rate on the impressions that do occur, because the ad unit reflects the same data quality issues visible to the algorithm.
The 5 Feed Issues That Destroy Shopping ROAS
| 1 | Poor title structureMost product titles in poorly maintained feeds are written for internal reference rather than for search matching. They prioritise SKU codes, internal category names, or brand shorthand that means something to the warehouse team and nothing to a search algorithm trying to match a consumer query. A title like ‘WM-4421-BLK-L’ tells Google almost nothing. A title like ‘Women’s Merino Wool Crew Neck Sweater Black Large’ tells Google exactly which queries to match. |
| 2 | Missing or incorrect GTINsGTINs (Global Trade Item Numbers, including EANs, UPCs, and ISBNs) are the unique identifiers that allow Google to match your products to manufacturer data and to aggregate reviews, pricing comparisons, and search behaviour at the product level. Missing GTINs reduce eligibility for Google’s enhanced product listings and can result in lower auction priority for products where GTINs are available from other merchants. For branded products, missing GTINs are almost always a feed management gap rather than a genuine data absence. |
| 3 | Generic product descriptionsDescriptions are weighted less heavily than titles in Google’s relevance algorithm, but they are not irrelevant. A description that consists of a single marketing sentence copied from the product homepage provides no additional keyword signal. A description that includes material composition, dimensions, use cases, compatible products, and care instructions provides a significant secondary signal layer that helps Google match the product to long-tail queries the title alone would not capture. |
| 4 | Incorrect product type taxonomyThe product type field is a merchant-defined categorisation that Google uses to understand the context of a product when its title and description are ambiguous. Errors here range from leaving the field empty, to using a taxonomy that makes sense internally but does not align with how Google’s category system organises similar products. Miscategorised products can surface in entirely wrong query contexts, generating irrelevant impressions that degrade CTR and signal to the algorithm that the product is a poor match for the queries it is appearing for. |
| 5 | Stale pricing and availability dataA feed that does not update frequently enough creates a mismatch between what the listing shows and what the landing page shows. A product listed at $79 in the feed that has been repriced to $95 on the site generates a policy violation and can result in product disapproval. A product showing as in-stock in the feed that is actually out of stock on the site generates a click to a dead end, with the CPC charged and zero conversion opportunity. Both scenarios are mechanical ROAS destroyers that have nothing to do with bid strategy. |
How to Audit Your Feed in 30 Minutes
A feed audit does not require specialist tooling to produce useful findings. The Merchant Center diagnostics tab surfaces the most critical issues automatically. A structured 30-minute review of those diagnostics, combined with a manual sample review of titles and descriptions, will identify the majority of performance-affecting problems in any feed.
| What to Check | Where to Look / What to Measure | Priority |
| Disapproval rate | Merchant Center > Diagnostics. Any disapproval rate above 2% requires immediate investigation. | Critical |
| Missing GTINs | Filter active products by GTIN field. Count products where field is empty or populated with placeholder values. | Critical |
| Title length distribution | Export feed, measure character count per title. Titles under 40 characters are almost certainly under-optimised. | High |
| Price and availability sync | Compare feed export timestamp to last site update. Any lag over 6 hours on a high-velocity catalogue is a risk. | High |
| Description word count | Filter descriptions under 100 words. These are candidates for enrichment regardless of other issues. | Medium |
| Product type field completion | Check what percentage of active SKUs have a populated product type field. Empty = lost categorisation signal. | Medium |
A disapproval rate above 5% in any product group warrants pausing that group until the underlying data issues are resolved. Running disapproved products in an active campaign wastes budget on impressions that cannot convert and sends negative quality signals to the algorithm.
The Optimisation Hierarchy: What to Fix First
Not all feed improvements return the same value. The hierarchy below reflects the typical impact-per-effort ratio across e-commerce programmes. Deviations from this sequence are occasionally justified by specific catalogue characteristics, but as a starting framework, it holds across most verticals.
Titles first. No other feed improvement delivers as much ROAS impact per unit of effort as title restructuring. The change is immediate: within the next feed refresh cycle, Google re-evaluates the keyword matching for every retitled product. For large catalogues, titles should be restructured systematically by category using a defined template (covered in depth in Article #08 of this series) rather than product by product.
GTINs second. For branded products where GTINs exist but are missing from the feed, adding them is a data quality fix with a predictable positive outcome. For unbranded or custom products where no GTIN exists, the exemption process in Merchant Center should be completed properly rather than leaving the field empty, which triggers a different set of warnings.
Descriptions third. Description enrichment delivers secondary keyword signal and supports long-tail query matching. The effort required is higher than title changes for large catalogues, so prioritise by product group: start with the product groups where titles have already been optimised and ROAS remains below target, as description quality is more likely to be the limiting factor in those cases.
Custom labels and segmentation last. Custom labels allow campaign segmentation by feed attribute: margin tier, seasonality, stock level, or promotional status. This is a powerful optimisation layer, but it requires the foundational data quality work above to be in place first. Custom labels on a feed with poor titles and missing GTINs produce finely segmented campaigns built on a broken data foundation.
Advanced Feed Signals That Move Performance
Once the foundational feed attributes are clean, a second tier of optimisations addresses more granular performance levers.
Promotional text and sale price attributes. Google surfaces sale indicators in Shopping ads when the sale price attribute is populated correctly alongside the original price. This visual treatment increases CTR on the listing independently of the bid, because the price comparison is visible before the click. Merchants running promotions who do not use the sale price attribute are leaving a CTR improvement on the table on every promotional period.
Custom labels for margin-based bidding. A common and high-value use of custom labels is segmenting products by gross margin tier. Products with high margins can sustain higher CPCs and lower ROAS targets. Products with thin margins need tighter CPA controls or should be excluded from competitive keyword auctions entirely. Without this segmentation, a target ROAS bid strategy applied at the campaign level blends high-margin and low-margin products and optimises toward an average that serves neither category well.
Feed rules for dynamic attribute generation. Most feed management platforms allow rule-based transformations to be applied at the feed processing stage: appending colour and size to titles automatically, generating descriptions from structured product attributes, or flagging out-of-stock products for bid suppression. These rules reduce the manual effort of feed maintenance at scale and ensure that new products entering the catalogue are processed consistently from the moment they are added.
Platforms like EnhanceFeed are built specifically for this layer of feed intelligence, using AI-driven attribute enrichment and rule-based transformation to maintain feed quality at catalogue scale without requiring manual intervention on every product update. For teams managing tens of thousands of SKUs across multiple channels, the difference between manual feed management and an intelligent feed layer is typically measured in percentage points of ROAS, not basis points.
Feed Quality Is Bid Quality
The framing that Shopping performance is primarily a bidding problem is understandable. Bids are the most visible control lever in the campaign interface, and the feedback loop between bid changes and performance changes is faster and more legible than the feedback loop between feed changes and performance changes.
But the underlying logic of Google’s Shopping auction is that relevance determines eligibility and bid determines position within relevance. A feed that is structurally poor limits the relevance score of every product in it, and no bid strategy operates above that ceiling. Improving the feed does not just improve one performance metric. It raises the ceiling for every subsequent optimisation.
The teams that consistently outperform on Shopping ROAS are not necessarily the ones with the most sophisticated bidding strategies. They are the ones who treat the feed as a performance asset that requires the same rigour as campaign structure and creative. The bids follow the data. Make sure the data is worth following.
