The feed is the campaign
In a search campaign the advertiser chooses keywords, writes ads and picks landing pages. In a shopping or catalogue campaign the platform does all three from the product feed: it decides which searches or audiences a product is relevant to from the title, description, category and attributes; it builds the ad from the image, title and price; and it sends the click to the product's link. The advertiser's influence runs through the data. A title that names the brand, product type, key attribute and size is matched to more of the right searches than one that repeats the internal product code; an image that shows the product clearly on a plain background is clicked more often than a lifestyle shot with the product small in the corner; a price that matches the page is allowed to show while a price that does not is disapproved.
This is why feed work is marketing work rather than a data task handed to IT. The people who know which attributes buyers search by, which products carry margin and which images sell are the people who should decide what the feed says. Google's Merchant Center Help documents the required and recommended attributes and the specification for each; the product feed guide in the Google Ads hub walks through them for Merchant Center, and the Meta catalogue ads guide covers the Meta equivalent. The Google Merchant Center expertise page describes how the author approaches feed work in practice. This guide stays above the platform: what every feed needs regardless of where it goes.
Four principles hold across platforms. Completeness: every attribute the platform documents as required, and as many recommended ones as the catalogue can supply, because each missing attribute is a match the product cannot make. Accuracy: price, availability and link must agree with the live page, because platforms check and disapprove on mismatch. Specificity: titles and descriptions written in the buyer's vocabulary, not the warehouse's. Segmentation: labels that let campaigns treat products differently by margin, season or performance, since a feed that cannot be split cannot be bid on intelligently.
Feed quality and disapprovals
Disapprovals are the visible failures; weak data is the invisible one. A product can be approved and still never show because its title matches nothing people search for. The table covers both.
| Problem | What it looks like | Why it costs | Fix |
|---|---|---|---|
| Missing identifiers | Products without gtin, brand or mpn where the platform expects them | Products with identifiers are matched to more searches and to the platform's catalogue entry; missing ones may be limited or disapproved | Source identifiers from the supplier; mark genuinely unbranded items the way the platform documents |
| Price mismatch | Feed price differs from the page, often after a promotion or currency change | Disapproval until corrected; lost impressions during the gap | Generate the feed from the live price; shorten the refresh interval; add structured data on the page so the platform can confirm |
| Availability mismatch | Feed says in stock, page says out of stock, or the reverse | Disapproval, or clicks to a product that cannot be bought | Sync availability from the inventory system at the same cadence as price |
| Weak titles | Internal codes, missing brand or attribute, all capitals | Fewer relevant matches; lower click-through where shown | Rewrite to brand, type, key attribute, variant, size, in that order, within the documented length |
| Poor images | Watermarks, text overlays, placeholder images, product too small | Disapproval under image policy, or low click-through | Plain-background primary image meeting the platform's documented requirements; lifestyle images as additional |
| Wrong category | Products mapped to a vague or incorrect category | Mismatched searches and audiences; wrong tax or policy treatment in some markets | Map to the most specific category in the platform's taxonomy |
| Policy violations | Restricted products, misleading claims, prohibited content in descriptions | Disapproval and, if repeated, account-level action | Read the stated reason; fix the cause; appeal only after fixing |
| Link problems | Redirects, tracking parameters that break, pages that differ by location | Disapproval for landing page issues; wasted clicks | Direct links to the canonical product page; tracking appended in the way the platform documents |
| Shipping and tax gaps | Missing delivery cost or time settings for a market | Products not eligible in that market; mismatched promises at checkout | Configure the documented delivery and tax settings per market and keep them current |
Both Google and Meta expose a diagnostics view that lists items with issues and the reason, and both document the reason codes in their help centres; the right habit is to read the stated reason rather than guess. A weekly review of the diagnostics page, with the fixes made in the source system rather than patched in the platform, keeps a catalogue healthy. The checker below runs the structural checks on a sample of feed rows: required attributes present, identifiers plausible, prices and availability in the expected format, titles within length. It does not replace the platform's own diagnostics, but it catches most problems before the feed is submitted.
Product feed checker
Google Shopping and Performance Max in brief
Google offers two campaign types that consume a Merchant Center feed. Standard Shopping shows product listings on search results and the Shopping tab, with the advertiser controlling which products are grouped together and what each group bids; it is the type that gives the most control and the clearest reporting, and most stores start with it. Performance Max takes the same feed and adds Google's other inventory (search, YouTube, Display, Discover, Gmail, Maps), choosing placements and bids automatically towards a conversion or value target; Google documents that it works from the feed, the asset groupsThe Performance Max unit holding the text, images, videos and signals for one theme, product set or audience. and the audience signals the advertiser supplies, and that listing groups decide which products each asset group may show.
The choice between them is about control and data. Standard Shopping lets the store bid differently by margin and read which searches produced which orders; Performance Max asks for a target and a feed and reports at a coarser level, which is reasonable once there is enough conversion history for the automation to learn from and the store has a way to separate products by margin. The Google Shopping and Performance Max hub covers the comparison and when to run both, and the Performance Max structure guide covers how many campaigns to run, listing groups, brand exclusions and reading its reports. Before scaling it alongside Standard Shopping, the Performance Max uplift test checks whether it grows total store revenue or only takes credit for orders the other campaigns would have won. Conversion tracking with order values is a precondition for either; the conversion tracking setup tutorial covers it. A store that starts with Standard Shopping builds its product groups, margin-based bids and campaign priorities with the Standard Shopping structure guide. A brand that is also sold through retailers can buy product ads on those retailers' own sites instead, and the comparison of retail media and Shopping ads sets out how the two differ in data, margin and measurement.
Meta catalogue ads in brief
Meta's catalogue ads use the same idea with a different matching engine. A catalogue is created in Meta's
commerce tools from a feed or a platform integration, product sets segment it, and ads are built from
templates that render each product's image, name and price. Two uses dominate. Dynamic retargeting shows
people the products they viewed, added to cart or bought, using the events the pixel or server-side
integration reports (ViewContent, AddToCart, Purchase) and the product identifiers those events
carry; Meta's help centre documents the event and parameter requirements. Catalogue prospecting shows
products to people who have not visited, with the delivery system choosing which product to show each
person.
Because nobody searched for anything, creative matters more than on Google: the template, the overlays, the intro card and the video that precedes the carousel decide whether the feed stops the scroll. The Meta catalogue ads guide covers product sets, retargeting windows, prospecting and creative; the Meta commerce hub covers catalogue setup and health, and the Meta pixel setup tutorial covers the events the retargeting depends on. For a store, the practical rule is that the Meta feed should be generated from the same source as the Google feed, with the attribute names Meta documents, so that a price change reaches both on the same schedule. When a store connects either feed for the first time, the Merchant Center feed setup checklist and the Meta catalog setup checklist are the one-off gates to pass before the first campaign uses it.
Margin-aware targets and break-even
Platforms optimise towards the target they are given, and most stores give them a revenue target: a return on ad spend (ROASRevenue generated per unit of ad spend.) that the campaign should clear. The target is only meaningful if it was derived from margin. Break-even return is the revenue a unit of ad spend must bring back for the order to cover its product, delivery and payment costs; a target below it loses money on every order the platform proudly reports. Because margin differs by product, so does the break-even return, and a single store-wide target quietly subsidises low-margin products with high-margin ones.
The method has three steps. Group products by gross margin band using custom labels in the feed. Calculate the break-even return for each band from its margin, then set the campaign target above it by the profit the store wants per order; the break-even calculator and the ROAS calculator do the arithmetic. Then read results by band, not by campaign, because a campaign that mixes bands hides which products are paying for the others. Where a platform reports cost per order (CPACost of each conversion.) instead, the same logic applies: the allowable cost per order is the gross margin on that order minus the profit wanted. Order value (AOVAverage revenue per order.) enters because a campaign that sells multi-item baskets can bear a higher cost per order than one that sells single items.
Two refinements matter for stores with repeat purchase. First, a product that brings customers who reorder can justify a first-order target below break-even, provided the repeat margin is measured rather than assumed; the e-commerce marketing hub sets out the equation. Second, returns are a cost: categories with high return ratesShare of orders or units sold that customers sent back. need a lower allowable cost per order than their shelf margin suggests.
Which bands deserve paid support in the first place, and what role each plays as an entry, hero, profit or clearance line, is a plan-level decision that the guide to choosing which products to put marketing behind makes before targets are set here.
Setting targets by margin band
A footwear store groups its catalogue into three margin bands with custom labels. Band A (own-brand trainers) carries 55% gross margin, so its break-even ROAS is about 1.8; the store wants £12 of profit per £90 order and sets a target of 2.4. Band B (branded boots) carries 32% margin, break-even 3.1, target 4.0. Band C (clearance) carries 18% margin, break-even 5.6, and the store decides not to advertise it at all. Before the split, one campaign ran at a blended 3.5 target and reported success while Band C sold at a loss and Band A was held back by a target it beat easily. After the split, Band A spend rises 40%, Band C spend goes to zero, and total profit after ads is higher on slightly lower revenue. A quarter later the store adds a 20% return rate to Band B's calculation and raises its target to 4.6.
Feeds across platforms: one source of truth
A store that advertises on Google, Meta and a marketplace or two ends up with several feeds, and the common failure is that they drift: a price updated in one, a title improved in another, availability synced to none. The remedy is one canonical product dataset, usually the store platform's own catalogue enriched with the attributes advertising needs, from which each platform's feed is generated in the format and attribute names that platform documents. Changes are made once, upstream; the feeds are outputs, not places to edit.
Three practices follow. Keep enrichment (better titles, margin labels, seasonal labels, lifestyle images) in a layer the store controls, whether a feed management tool or a set of rules on the export, so that it survives a platform migration. Refresh at the cadence of the data that changes fastest, which is usually availability and price; a daily feed for a store that sells out hourly disapproves itself. And treat structured data on the product page as part of the same system: platforms compare the feed to the page, and a page that carries the same price and availability in markup, as the structured data guide describes, lets them confirm it automatically.
The product page itself has to do its job once the click arrives, which is the subject of the product page optimisation guide, and the wider acquisition picture, including whether a marketplace belongs in the mix at all, sits in the acquisition hub and the marketplaces comparison.
Feed work belongs in a plan's Reach column, and RACE for ecommerce sales shows it sitting beside media spend, product-page, checkout and repeat-order actions in a seasonal plan.
For a Google-specific configuration, the recommended Google Ads setup for ecommerce sales covers Shopping and Performance Max roles, margin-banded product groups and a target ROAS set from break-even.
For sponsored product ads run inside a marketplace rather than a shopping feed, the marketplace sellers' ROAS calculator guide explains entering post-fee margin and converting the return to ACoSMarketplace advertising cost divided by the sales the ad console attributes to ads, shown as a percentage; the inverse of ROAS..
Where these ads serve an online sales target, the online sales objective hub places them among the other channels that sell, and the paid search channel page for online sales sets the margin-based targets and query segments for the search side.
When Google limits items because their identifiers are missing or invalid, the Merchant Center missing identifiers diagnosis traces each group of items back to the store record, the supplier file or the export that lost the codes.
Common mistakes
- Treating the feed as an export. It is the targeting, the creative and the landing page; it deserves the attention a search campaign's keywords get.
- One store-wide ROAS target. Low-margin products lose money while high-margin ones are starved.
- Fixing disapprovals in the platform. Patches made in a feed rule vanish on the next export; fix the source.
- Titles written for the warehouse. Product codes and internal names match nothing people type.
- Letting price and availability drift from the page. The most common disapproval, and the easiest to prevent with a shorter refresh.
- Ignoring returns. A category with a high return rate has a lower real margin than its target assumes.
- Different feeds for different platforms, edited separately. One source, many outputs.
- Judging Performance Max on its own reports alone. Read margin by product band outside the platform before raising the budget.
