Here's the fact that should reorganise how you spend your time on Shopping campaigns: you don't choose keywords. There's no keyword field, no match type, no negative list doing the heavy lifting. The platform reads your product data and decides which searches your items appear for. Which means your product feed is doing the job keywords do in a Search campaign — it is your targeting. Once that lands, the usual advice to "tidy up your feed" stops sounding like housekeeping and starts sounding like what it actually is: the highest-leverage work available in ecommerce advertising.
What the feed actually controls
Your feed does three jobs at once, and most guides only address the third.
1. Matching. Your titles, attributes and categories determine which searches you're eligible to appear for. This is your targeting.
2. Eligibility. Accurate prices, availability, identifiers and policy compliance determine whether the product can serve at all. Fail here and nothing else matters.
3. Presentation. Your image, title and price are the ad. There's no separate creative to write.
Most feed advice covers presentation. The money is in the first two.
That framing also explains why feed work has become more valuable, not less, as campaigns automated. When bidding, placements and granular targeting moved out of your hands — the shift we set out in what's left to control in PPC — the quality of the data you supply became the main remaining lever. In feed-driven campaigns, the feed is the account.
The reframe In Shopping you don't pick keywords — your product data picks them for you. The feed isn't admin. It's the targeting layer.
The title: your highest-value field
If you optimise one thing, optimise titles. They carry the most matching weight and they're the most visible part of the ad, yet most feeds inherit whatever the ecommerce platform generated — often an internal product name written for a merchandiser rather than a shopper.
The principle is simple: include the words people actually search, and put them early because titles truncate. A structure that works across most categories runs brand, then product type, then the attributes that distinguish this item — colour, size, material, model, capacity. What belongs in that attribute slot varies by vertical, which is where generic advice falls down.
| Category | Structure that tends to work |
|---|---|
| Apparel | Brand + gender + product type + attributes + size + colour |
| Electronics | Brand + model number + product type + key specification |
| Home & furniture | Brand + product type + material + dimensions + colour |
| Consumables | Brand + product type + variant + size or count |
| Parts & accessories | Brand + product type + compatibility (fits X) + specification |
Two rules apply regardless of category. Use the language shoppers use, not your internal naming — if customers search "trainers" and your feed says "athletic footwear," you're invisible to them, which is the same intent-matching discipline behind sound keyword research. And keep promotional language out: "SALE," "FREE SHIPPING" and all-caps don't help matching, look worse, and can breach policy.
The attributes that decide matching
Beyond titles, a handful of structured fields do the real work of telling the platform what your product is. Product identifiers such as GTINs let the system recognise the exact item and its market context, and missing them can restrict eligibility outright. Brand, product category and your own product type field refine which queries you're a candidate for.
Then there are the descriptive attributes — colour, size, material, pattern, age group, gender, condition — which look like optional metadata and function as targeting refinements. A shopper searching "navy wool overcoat mens large" is filtering on four attributes; if any are missing from your data, you can't be the answer. Populate every field that genuinely applies to your product, and leave blank the ones that don't rather than guessing.
Descriptions matter less than titles for matching but still contribute, and they're worth writing properly rather than dumping raw manufacturer copy. The same specificity that helps a product page convert helps here — the principles in product page optimisation transfer almost directly.
PixelMe
Drives and attributes external ad traffic to Amazon listings with branded links and retargeting.
Best for: Amazon sellers driving external ad traffic & retargeting
The silent revenue killer: disapprovals
Feed errors are uniquely dangerous because they're quiet. A disapproved product doesn't trigger an alarm — it simply stops appearing, and revenue drops without an obvious cause. Teams routinely discover weeks later that a chunk of their catalogue went dark after a price update.
The most common causes are mismatches between feed and website, particularly price and availability differing from what the landing page shows. Others include missing identifiers, image problems, wrong category assignment, and policy issues on restricted items. The fix is process rather than cleverness: check your diagnostics on a fixed schedule, treat disapprovals as a revenue incident rather than a technical ticket, and make sure feed updates keep pace with site changes — daily or better if your prices and stock move frequently.
Custom labels: where profit enters the picture
Now the part that separates a feed optimised for performance from one optimised for profit — and the part most guides skip entirely.
The platform knows what your products cost the customer. It has no idea what they cost you. Left to itself, automated bidding optimises toward revenue, which means it will happily pour budget into your highest-volume, lowest-margin lines and report an excellent return while your profit goes sideways. It's doing exactly what it was told.
Custom labels are the fix: optional fields you define yourself, which is how business context gets into a system that otherwise sees only products and prices. Use them to tag margin bands, bestsellers, seasonal items, clearance stock, price tiers, and new arrivals — then segment and set targets accordingly, so a 60%-margin product can be bid harder than a 8%-margin one. This is the feed-level expression of the broader principle that automated systems pursue whatever you tell them to value, which is why it belongs alongside conversion values you can trust as foundational rather than advanced work.
| Visibility-focused | Profit-focused |
|---|---|
| Titles stuffed for maximum matching | Titles matched to buying-intent queries |
| All products treated equally | Margin bands in custom labels driving targets |
| Optimised to ROAS | Optimised to contribution after cost of goods |
| Every SKU in the feed | Loss-makers and unavailable lines excluded |
| Success = impressions and clicks | Success = profitable orders |
The exclusion row deserves emphasis, because it's counterintuitive. Not every product belongs in your feed. Items with negative contribution after shipping and returns, perpetually out-of-stock lines, and products with terrible conversion rates all consume budget and drag your averages down. Removing them usually improves performance immediately — one of the rare optimisations that costs nothing and takes an afternoon.
Images and presentation
Since your image is your ad, it carries disproportionate weight in click-through. The essentials are unglamorous: high resolution, the product filling most of the frame, a clean background, accurate colour, and consistency across your catalogue so your results don't look chaotic next to competitors. Additional images give the platform more to work with in different placements.
Two mistakes recur. Promotional overlays and watermarks on the main image typically breach policy and get products disapproved. And lifestyle-only shots, while beautiful, often underperform against clean product shots in a grid where shoppers are comparing items at a glance.
Feeding automation properly
In automated campaign types, the feed carries even more weight because there's less else to adjust. Practically, that means a few habits matter more than they used to. Update frequently — real-time or near-real-time price and stock updates prevent both disapprovals and wasted spend on unavailable items. Supply complete data, since every populated field is another way the system can match and present your products. And use your labels to guide the automation toward the products you actually want to sell, rather than hoping it works that out.
The relationship between feed quality and campaign structure is worth naming too: strong feed data reduces how much structural segmentation you need, which suits the consolidation logic in structuring an account for scale and control. Better data means you can run fewer, better-fed campaigns.
A maintenance rhythm
Feed optimisation isn't a project you finish. Prices change, stock moves, new products arrive, and platforms adjust requirements. A workable cadence: check diagnostics weekly for disapprovals and errors; review search terms monthly to see which queries your products are matching and adjust titles where the matching is wrong; and revisit margin labels quarterly, or whenever costs change, so your profit data doesn't drift out of date.
That last one is the most commonly neglected. A margin label set eighteen months ago, before supplier costs rose, is now actively misleading your bidding — the system is optimising confidently toward yesterday's economics.
Why this matters more in 2026
Two pressures make feed quality more consequential than it was. Media costs have risen substantially, as we cover in rising CPMs and how advertisers are adapting, so wasted impressions on badly matched products are more expensive than they used to be. And structured product data increasingly feeds shopping experiences beyond traditional Shopping ads — including AI-assisted product discovery, the shift covered in how AI search is changing product discovery. Clean, complete, accurate product data is becoming the substrate for more than one channel, which raises the return on getting it right once.
And of course the feed only gets someone to your site. What happens next is a separate problem — one worth solving in parallel, since sending well-matched traffic into a leaky checkout flow simply buys expensive abandonment. Getting both right is where an experienced ecommerce marketing team tends to earn back its fee fastest.
The bottom line
Your product feed isn't a technical prerequisite for running Shopping ads — it's the mechanism that decides which searches you appear for, whether you're eligible to appear at all, and how you look when you do. Rewrite titles so they lead with what shoppers search rather than what your system called the product, populate every attribute that genuinely applies, and treat disapprovals as lost revenue rather than admin. Then take the step most advertisers skip: put your margin data into custom labels, so automated bidding optimises for profit rather than volume, and remove the products that lose money on every sale. Campaign settings are increasingly out of your hands. The feed still isn't.
Is your product feed working for profit or just for volume?
We rebuild feeds around buying intent and real margins — so your Shopping spend grows contribution, not just revenue.
Explore Performance Marketing →Frequently asked questions
What is product feed optimization?
Product feed optimization is the practice of improving the product data you send to shopping platforms so your items match the right searches, stay eligible to serve, and present well against competitors. It covers titles, descriptions, product identifiers, categories, images, pricing and availability accuracy, and supplementary fields like custom labels. It matters more than campaign settings in Shopping advertising because you don't select keywords — the platform matches queries to your product data, which makes that data your targeting mechanism.
What makes a good Shopping ad product title?
Front-load the terms shoppers actually search and include the attributes that distinguish the product. A workable pattern is brand, then product type, then key attributes such as colour, size, material or model, arranged so the important words appear early since titles truncate. Match the structure to the category — apparel benefits from gender, size and colour, while electronics benefit from model numbers and specifications. Avoid promotional language, all-caps and internal jargon, none of which help matching and some of which breach policy.
Why are my products disapproved in Merchant Center?
The most common causes are mismatches between your feed and your website — usually price or availability differing from the landing page — along with missing required identifiers such as GTINs, image problems, incorrect category assignment, and policy issues with restricted products. Disapprovals are especially costly because they're silent: nothing alerts you that revenue has stopped, the products simply stop serving. Checking the diagnostics area regularly is the only reliable way to catch them promptly.
What are custom labels used for in a product feed?
Custom labels are optional fields you define yourself, and they're how you get business context into a system that otherwise only sees products and prices. Common uses include tagging margin bands, bestsellers, seasonal items, clearance stock, price tiers, and new arrivals. Their real value is enabling profit-aware bidding: without margin data, automated bidding optimises toward revenue and will happily scale your least profitable products. Custom labels let you segment and bid according to what each product is actually worth to you.
Does the product feed matter for Performance Max campaigns?
It matters more, not less. As automation absorbed the manual controls advertisers used to rely on — bidding, placements, granular targeting — the quality of the data you supply became the main remaining lever. In feed-driven campaigns your product data determines what the system can match, how it presents your products, and which items it chooses to push. A campaign built on a weak feed will underperform regardless of how the settings are configured, because the settings are no longer where the leverage is.
