Est. reading time: 7 minutes

Catalog ads earn a reputation as the set-and-forget corner of a Meta account. Connect the store, pick a product set, let the system decide who sees what. When performance sags, the fixes all happen in Ads Manager, new audiences, new budget splits, new copy around the same catalog.

What rarely gets opened is the feed itself. Every catalog ad impression is assembled at delivery from feed fields, the image, the title, the price, the availability flag, and in most accounts nobody has reviewed those fields since the store launched. The static ads get a creative review before a dollar touches them. The format built from the feed never does, and it shows.

The account-level version of this discipline is what we actually look for when we audit a Meta ad account. This is the product-layer version, the places where a feed quietly caps what the campaigns built on it can do.

Catalog ads assemble the creative from your feed at delivery time

Advantage+ catalog ads (the format Meta used to call dynamic ads) don’t run a finished creative. They build one per impression, pulling the image, title, price, and availability from the catalog and arranging them into whatever the placement calls for. Your pixel and Conversions API data decide which product a person sees. The feed decides everything about how that product looks when they see it.

That split explains the pattern. Teams pour review cycles into static and video ads, then hand a large share of the budget to a format whose creative inputs are whatever the platform export happened to contain. The fix starts with looking at the output. Pull up your ten best sellers, preview them as catalog ads in feed and Stories placements, and read them the way a shopper would. Feeds that have never been audited rarely survive that reading.

Titles written for the product page, truncated by the placement

Feed titles inherit whatever the store uses, which means SEO strings, brand-first ordering, and variant suffixes riding along (“Heather Gray / M” does nothing for a stranger in a feed). Placements then cut titles short, so anything after the first few words may never render. The title’s job in a catalog ad is closer to a headline than a label.

Front-load the part that sells. Product type and the differentiating attribute first, brand last unless the brand is the draw, variant data stripped out entirely. Do the rewriting at the source where you can, and with your feed tool’s rules where you can’t. While you’re in there, populate sale_price alongside price, since a markdown the feed doesn’t know about is a markdown the ad can’t show.

Images shot for a white background, cropped into a square

Catalog placements render mostly square in feeds and carousels and vertical in Stories, and Meta crops your image to fit. The default packshot, a small product floating in white space at whatever ratio the photographer delivered, turns into a thumbnail where the product occupies a quarter of the tile. Text baked into the image fares worse, because the crop can land anywhere.

Meta’s image specifications for catalogs set the floor at 500 by 500 pixels, and images near the floor look soft on modern screens, so treat 1024 by 1024 as the working minimum. Frame the product to fill a square, keep it legible at thumbnail size, and use the feed’s additional image slots for a lifestyle alternative where the category sells on context. The image is the entire first impression in this format. It deserves the same standard as any ad you’d approve by hand.

When pixel IDs and catalog IDs disagree, retargeting goes generic

Dynamic retargeting works on a join. The pixel and Conversions API events behind it (ViewContent, AddToCart, Purchase) carry a content_ids parameter, and Meta matches those IDs against the id field in your feed, character for character. When they match, the person who viewed a product gets an ad for that product. When they don’t, the system falls back to guessing, and the retargeting you built the catalog for goes generic.

Shopify accounts break this join in predictable ways. The native integration sends variant-level IDs while a third-party feed carries product-level IDs, or the pixel and the Conversions API disagree with each other, and no error fires anywhere. The symptom is quieter, retargeting that serves popular products instead of viewed ones and performance that never justifies the setup. Events Manager surfaces a catalog match rate for the pixel attached to your catalog, and that number is the diagnostic. If it’s low, align the IDs, either by letting one integration own both the events and the feed or by setting the catalog to expect the ID format your events actually send.

The audience side of this is territory we’ve covered, in mastering Facebook ad retargeting for ecommerce and how to use purchase data to build better retargeting audiences. The feed side is what makes those audiences worth building.

A stale feed spends money on products you can’t sell

Availability and price live in the feed, and they age the moment the upload finishes. A daily fetch means a 9 a.m. sellout keeps advertising until tomorrow, paying for clicks into an out-of-stock page. A price change that hasn’t synced means the ad quotes one number and the site charges another, which is the kind of mismatch that ends sessions and, repeated at scale, runs against Meta’s expectations for catalog accuracy.

Match the sync method to how fast your store moves. The platform integrations keep catalogs updated automatically, and hosted feeds can be fetched on a schedule as tight as hourly. Whatever the method, keep one data source per catalog and update it in place rather than stacking new uploads, and make the Data sources tab in Commerce Manager a standing check, since fetch errors and item rejections land there quietly.

Custom labels turn one catalog into deliberate campaigns

Left alone, most accounts run catalog campaigns against an all-products set, and delivery concentrates wherever conversions come cheapest. That’s frequently the discounted line or the low-margin bestseller, which is how a campaign hits its ROAS target while contribution margin goes sideways. The system optimizes what it’s given. An all-products set gives it no instructions.

The feed carries five custom label fields (custom_label_0 through 4) that exist for exactly this. Label items by margin tier, seasonality, stock depth, price band, or hero status, then build product sets from the labels and point campaigns at the sets. Prospecting runs on the products with the margin to pay for cold traffic, retargeting stays broad, and low-stock items get excluded before budget chases inventory that can’t support a sale. None of this is available from inside Ads Manager. It’s feed work, and that’s the quiet theme of this whole list. The fix sits upstream, in a file most teams never open.

None of these fixes requires a bigger budget or a new campaign type. They require someone to own the feed the way someone already owns the ads, because in this format the feed is where the ads come from. Run the preview exercise, check the match rate, and read your own titles as a stranger would. If you’d rather have company for that pass, it’s the kind of audit we like doing.