Optimise the feed. You have read that advice a dozen times, and not one told you what to change. Your products are approved, your bids look sane, yet your ROAS sits flat. So which part of your product data is Google punishing?
There is a version of your product feed that Google reads differently from the one you think you uploaded. Every attribute looks present in your spreadsheet. In practice, three or four fields are working against your campaigns, and nothing in your Google Ads account will point at them. You see the symptom as a flat ROAS, a rising CPC, or a product that quietly stops appearing for queries it used to own. That is not a bidding problem, and raising your target ROAS will not fix it. What fixes it is knowing which fields Google rewards, which ones it demotes, and in what order to deal with them — and that order is not the order most guides give you.
What a Product Feed Actually Is (and Why It Is Your Ad)
Google Shopping has no ad copy for you to write and no keywords for you to bid on. Your product data is the ad. Everything Google knows about what you sell, who should see it, and how much it is worth showing arrives through one channel: the structured product data you send to Google Merchant Center.
That data powers two things at once. It decides eligibility — whether a product can be shown at all — and it decides matching — which search queries your product becomes a candidate for. A bid can only compete among products that already passed both filters.
Quick Answer: Feed optimization means editing your Merchant Center product data so Google can match your products to the right queries, classify them correctly, and keep them eligible to serve. It is not a bidding task, and it is not a one-time setup. It happens in four places: approval status, identifiers, titles, and images.
If you are still building your first campaign, our guide to Google Shopping ads covers the setup and account structure. This article assumes that part is done and you want the feed to start paying for itself.
Here is the honest measure of the opportunity. Merchant Center accepts more than 60 product attributes. Most stores actively manage fewer than 15. That gap is where your feed work lives — not in the attributes you already filled in.
Over 60. Most stores actively manage fewer than 15. The gap between those two numbers is where the optimization opportunity lives.
The Attributes That Actually Move ROAS
Every feed guide hands you a list. A list is not a priority. Below is the order that matters, with the consequence of getting each one wrong — because "suboptimal" and "invisible" are very different problems.
Attribute | What happens when it is wrong | What you fix |
|---|---|---|
Approval status | Product is not eligible. Not approved means invisible; Limited means served only sometimes | Fix the error in Needs attention, then re-review |
GTIN, brand, MPN, identifier_exists | Disapproval, or approved but matched to the wrong queries | Supply real identifiers, or declare identifier_exists: no honestly |
Title | Google matches your ad to the wrong searches, or rewrites your title for you | Brand + product type + key attributes in the first 70 characters |
Image and image_link | Demotion, or a rejected image. Weak images lose the Shopping slot | 1500 x 1500 px, product fills 75-90% of the frame |
Price and availability | A mismatch between the feed and your landing page triggers Not approved | Keep price and stock in sync with the live product page |
custom_label_0 to 4 | Nothing breaks — you just cannot segment by margin, season or performance | Label products by margin, seasonality and top sellers |
Titles: the single most impactful attribute
Google uses your product title to decide which search queries trigger your ad. Not your description, not your ad group, your title.
The mechanics are strict. You have 150 characters, and the first 70 do the heavy lifting — brand, product type and key attributes belong there. Google prohibits ALL CAPS, promotional language such as "FREE SHIPPING" or "SALE", gimmicky characters, and repeating your store name when the brand attribute already carries it. Break those rules and you get one of two outcomes: disapproval, or Google silently rewriting your title for you through Product Data Customization. Letting Google write your titles is not optimization. It is outsourcing the one attribute that decides your matching.
Identifiers: where approved products still lose
Missing identifiers rarely give you a clean error. They give you a demotion — a product that is approved, technically serving, and matched to a much weaker set of queries than it deserves.
Supply the GTIN where one exists. Where a product genuinely has no GTIN, set identifier_exists to no honestly and make sure brand and mpn are filled. Products with GTINs get up to 40% more clicks, a figure attributed to Google through secondary sources rather than a single published benchmark — treat it as directional. The point stands: identifiers are what let Google connect your product to the same product everywhere else on the web, and that connection is what wins the query.
Images: a deadline worth diarizing
Images carry more weight than most retailers assume, and the rules are about to get stricter. From January 2027 the minimum image size rises to 500 x 500 px across all product categories. Aim higher: 1500 x 1500 px is the recommended working standard, with the product filling 75-90% of the frame.
One failure mode has nothing to do with image quality at all. If your robots.txt blocks Googlebot or Googlebot-image, Google cannot fetch the image and the product suffers regardless of how good the file is. It is a common and quiet cause of image load failures.
There is a detail almost nobody tells you: lifestyle_image_link feeds Display, YouTube and Performance Max creative — not standard Shopping. If you have been expecting lifestyle shots to change your Shopping results, that is why they have not.
Price, availability and the Limited status you are not checking
The usual summary of feed problems is "disapproved means invisible". True, but incomplete. Google's own documentation names five product statuses: Approved, Under review, Processing, Limited, and Not approved.
Limited is the one to care about. The product is showing on Google, but only in some instances, or with part of its data missing — eligible in some countries but not others, or not serving for some marketing methods. A partially served product looks healthy in a headline report and quietly leaks your ROAS every day. Most merchants never open the Visibility column to look for it.
Price and availability drive hard failures. A price that does not match your landing page, or a product marked in stock that is not, moves the item toward Not approved. Since 15 January 2026 the return_method attribute is required as well, with progressive disapproval following. And hidden products auto-archive after 14 days in Merchant Center — which is why a temporary stock-out can cost you an item permanently if nobody brings it back.
Custom labels: the segmentation you are leaving on the table
custom_label_0 through custom_label_4 are five free-text slots that do not affect Google's own rating of your product at all. That is exactly why they are useful. They let you segment products by margin, seasonality, inventory depth or past performance, and then bid differently on each segment.
Without labels you are bidding on everything at once and letting a low-margin accessory compete with your best seller. Set the labels once, then use them to build the product groups that make your campaign structure mean something.
Your Merchant Center Vocabulary Is Out of Date
Here is the problem with almost every guide you have read on this topic, including some published this year. They send you to interfaces that no longer exist.
Merchant Center Next became the default, and all retailers were migrated by September 2024. Google has been steadily removing the word feed from its own product. Match what you read online against what you actually see in your account:
What guides still call it | What Merchant Center calls it now |
|---|---|
feeds | data sources |
feed rules | Attribute Rules |
supplemental feeds | supplemental sources |
Diagnostics tab | Needs attention |
Active / Pending / Disapproved / Expiring | Approved / Under review / Processing / Limited / Not approved |
If you have been searching your account for a Diagnostics tab and concluding you are on the wrong plan, you were not. You were following a 2024 instruction. Practitioners following the change closely report that even the feed label is being retired toward "source label" as Google finishes deleting the old vocabulary.
The practical paths you need:
- Attribute Rules → Settings → Data Sources → Primary Sources → click the data source → Attribute Rules. Turn on the Advanced Data Source Management add-on first (Settings → Add-Ons).
- Supplemental sources → Settings → Data Sources → Supplemental Sources → Add Supplemental Product Data.
- Errors and warnings → the Needs attention tab, where both problems and their fixes surface. Approval state lives in the Status column; actual serving behaviour lives in the Visibility column. Read them as two different questions.
Attribute Rules vs Supplemental Sources
This decision confuses more merchants than any other part of Merchant Center, because the two features look interchangeable and are not.
Both override your primary data source without you having to edit the original file. Both are worth knowing. Use them for different jobs.
Use Attribute Rules when you are transforming an attribute at the source. They are built for attribute-level changes inside Merchant Center — renaming a value, appending text to a title, dropping a bad character. The honest limitation is scale: at thousands of SKUs, rules become genuinely unwieldy to maintain, and each one is another thing that can silently break.
Use a supplemental source when the override is data you maintain outside Merchant Center. A supplemental source is a separate file that layers over the primary one. Two constraints matter. A supplemental source cannot add new products — it only overrides attributes on products that already exist. And it is the wrong tool for patching one or two items; that is a job for a rule or for fixing the source.
Where several supplemental sources are in play, they run as a user-ordered stack, and the one placed highest wins. Order of application is thinly documented by Google, so treat that precedence carefully rather than assuming the behaviour.
Fix the source when the error is in the source. This is the step most teams skip. If your product data platform is exporting a truncated title or a stale price, a rule that patches it is technical debt. You now have two systems writing one attribute and no clear owner of the truth.
Feed Optimization for Performance Max vs Standard Shopping
The same feed drives both campaign types. It does not fail the same way in each, and that difference decides where your attention should go first.
In Standard Shopping, your feed determines which products are eligible and how well they match. You still have product groups, manual structure and negative keywords to compensate with. A weak attribute set narrows your reach; it does not blind the campaign entirely.
In Performance Max, the feed becomes the product supply for the listing groups inside your asset groups. There is no manual targeting layer to compensate for a thin attribute set, because the feed effectively is the targeting. A weak feed in PMax has nothing standing behind it.
That asymmetry is the reason feed health matters more in 2026 than it did two years ago. Performance Max grew from 25.9% to 33.1% of budgets across 57 retail accounts in one agency study, while Shopping spend fell from 21.3% to 12.4%. Those figures come from a single third-party study distributed as a press release, so treat them as directional — and note they measure budgets inside those accounts, not retail search ad spend overall, which is why they do not contradict Shopping's share of total retail spend.
The feed also now gates AI-era visibility. AI Mode shows roughly 95% fewer product listings than classic results and returns products for 23% of queries against 88% for classic results — again a third-party figure, but the direction is consistent with everything else. Structured data correlates with being cited in AI Overviews at roughly 3.1x, and 71% of ChatGPT-cited pages carry structured data. A product feed is structured data at commerce scale. Getting it right is now both a Shopping prerequisite and, increasingly, an AI-visibility one.
For the campaign-side view of that shift, see our guide to Performance Max campaigns and Performance Max management.
A Prioritized Feed Audit: Five Steps in Impact Order
Run these in order. Fixing titles before you have fixed identifiers is wasted effort, because a disapproved product does not care how good its title is.
1. Clear every disapproval and every Limited product. Open Needs attention. Sort by approval status first, then filter Visibility for products that are not fully serving. This is the highest-impact step and the only one that can restore products to eligibility. If diagnostics are opaque, our notes on Google Ads reporting cover how to read the account side of the same picture.
2. Fix identifiers. GTIN where one exists, brand and mpn always, identifier_exists set honestly. This is the step that turns approved-but-ignored products into products that match the queries they deserve.
3. Rewrite titles to the 150-character format. Brand + product type + key attributes in the first 70 characters. Remove promotional words and ALL CAPS. If a product has a model number, include it — shoppers searching a specific model convert at a far higher rate than generic queries. When you build title conventions, keep them consistent with how you handle Google Ads keywords, because the two systems are matching against the same searches.
4. Replace weak images. Hit the January 2027 standard early: 500 x 500 px minimum, 1500 x 1500 px target, product filling most of the frame. Confirm Googlebot can actually fetch your images.
5. Set your five custom labels. Margin on one, seasonality on another, best-sellers on a third. This is the step that makes everything above measurable, because it gives you the segments to report on.
Notice what is not on that list. Nothing here is a bidding change. Your feed decides whether a product is eligible and how well it can be matched, before any bid matters.
And notice how much of it is mechanical. The work is real, but it is repetitive, and it is exactly the kind of repetitive work that gets deferred indefinitely. This is where doing it by hand stops scaling: the difference between a store with 40 SKUs and a store with 4,000 is not the checklist, it is whether anyone can hold the whole feed in their head. That is the gap Google Shopping campaign management is built to close — feed-health flags, product-group insights and ROAS reporting in one place, so a feed problem surfaces as a flag instead of as a quarter of flat performance.
Your Feed Decides the Ceiling Your Bid Works Inside
Go back to the question you could not answer at the start. Which part of your product data is Google punishing? You now have a way to find out — not by guessing, but by walking the five steps in order and watching which one changes your visibility.
Start with Needs attention, because everything downstream depends on your products being eligible at all. Then confirm your identifiers, rewrite your titles to the format, upgrade your images before January 2027 makes it mandatory, and label your products so you can finally see which segment is paying for the rest.
None of it is glamorous. All of it compounds. The feed sets the ceiling, and your bid only decides where you land underneath it.
Frequently Asked Questions
What is Google Shopping feed optimization?
Feed optimization is the work of editing your Merchant Center product data so Google can match your products to the right searches, classify them correctly and keep them eligible to serve. It covers approval status, identifiers, titles, images, price and availability, and segmentation labels. It is separate from bidding — a perfect bid on a disapproved product achieves nothing.
Do I need a feed management tool, or can I do this manually?
Under roughly 500 SKUs, you can manage it yourself in Merchant Center using Attribute Rules or a supplemental source. Past that, the maintenance load becomes the problem rather than the knowledge. The tool question is really a question about whether one person can keep every attribute correct across every product, every week. If the answer is no, manual management is already failing quietly.
How often should I update my product feed?
Your primary data source should refresh at least daily for price and availability, because a price mismatch or a stock error moves a product toward Not approved. Then review Needs attention on a fixed weekly cadence. Do not only look for errors — check Visibility for Limited products too, since partial serving produces no error at all.
Why are my products disapproved in Merchant Center?
Almost always one of three causes: a required attribute is missing or invalid, the price or availability in the feed does not match your landing page, or the image fails Google's requirements or cannot be fetched. Google's own documentation notes reviews can take up to seven business days, and hidden products auto-archive after 14 days, so fix errors the day they appear rather than the week after.
Does feed optimization help Performance Max?
More than it helps Standard Shopping, in relative terms. In Performance Max your feed supplies the listing groups inside your asset groups, so there is no manual targeting layer to compensate for weak product data. A feed problem that costs you some reach in Standard Shopping costs you the campaign's foundation in PMax. Same file, higher stakes.
Your Shopping feed decides your ROAS before your bid does
Connect your Google Ads account and get feed-health flags, product-group insights and ROAS reporting in one place. Start free at studio.allable.ai.