Setting up Meta catalog ads from a clean product feed

Catalog ads assemble themselves from your product data, which means the campaign is only as good as the feed and the events behind it. Three things have to agree before anything works: the catalog holds accurate items, the site sends events describing what a shopper viewed or bought, and the product identifiers in those events match the identifiers in the catalog exactly. When retargeting shows the wrong item or shows nothing at all, the cause is almost always the third one.

Build in that order. Campaign settings come last and take the least time.

Get the catalog in by the route you can maintain

Most stores have three options: a platform sales channel that syncs the catalog automatically, a scheduled feed file, or a direct integration through an app. The right choice is the one that stays current without a person remembering to run it, since a catalog that lags the store by a day will advertise sold-out items and stale prices.

Whichever route you use, the catalog needs a scheduled refresh frequent enough for your stock movement, and inventory and price changes need to reach it before shoppers see them. The platform-specific connection steps are covered in how to connect a Shopify catalog to Meta.

If the store already maintains a shopping feed built to the Google product data specification, that file is a reasonable starting point because the field vocabulary overlaps heavily. Do not assume it maps perfectly. Check the required fields on the receiving side and fill the gaps with a supplemental file or a feed rule rather than editing the catalog, using the approach in feed rules without catalog changes.

Make the identifiers match, end to end

This is the step that decides whether dynamic retargeting works.

The identifier in the catalog must be the same string the site reports in its view, add to cart and purchase events. On platforms where products have both a parent identifier and per-variant identifiers, pick one convention and apply it everywhere. Mixing them produces a system that recognizes some products and silently ignores others.

A checklist:

• Decide whether the catalog is keyed on variant identifiers or parent identifiers, and document the choice.
• Confirm the site’s events send that same key, in the same format, including any prefix the sync applies.
• Group variants of one product under a shared item group id so size and color options are treated as one product with several options.
• Include GTINs where they exist, since identifiers improve matching across surfaces that use them.
• Verify with a real product: view it on the site, then check that the event payload contains an id present in the catalog.

Do this check on three products from different templates. A store with a custom product template for bundles frequently sends a different id shape from its standard products.

The feed fields that decide how the ad looks

Catalog ads render from feed data, so a field that looks like metadata is really creative.

Field	What it controls	What to watch
Title	The headline in most placements	Front-load the product identity; long titles truncate in feed placements
Description	Supporting text where the placement shows it	Write for a shopper, not for a search engine
Image link	The visual, cropped by the placement	Square-safe composition, since vertical and square crops cut the edges
Additional images	Alternate creative in some formats	Keep the same aspect ratio discipline
Price and sale price	The displayed price and any strike-through	Must match the landing page at the moment of the click
Availability	Whether the item can be shown	Stale availability advertises what you cannot sell
Brand	Matching and grouping	Use the manufacturer’s brand, consistently
Custom labels	Segmentation for product sets	Populate from margin, stock depth and lifecycle

Titles need a different shape from search-driven titles, because the audience is being interrupted rather than searching. Lead with the product identity and the one attribute that makes it recognizable, and keep the keyword stacking out of it. The category-by-category patterns are in product title formulas by category.

Build product sets from economics, not from the site menu

Product sets are the unit campaigns target, and the useful splits are financial rather than navigational:

• Core margin set. Products with healthy contribution and stock depth. Most spend goes here.
• Low margin or clearance set. Advertised at a different target, or excluded.
• New arrivals set, rebuilt automatically from a date field so it stays current.
• Hero and best seller set, for prospecting where recognition matters.
• Excluded set, covering restricted items, made-to-order products with long lead times and anything policy-sensitive.

Drive the sets from custom labels in the feed rather than from hand-picked lists, so a new product joins the right set on the day it is created. Hand-maintained sets decay within a quarter.

Event quality is half the system

Retargeting sets need accurate signals about what each visitor did, and purchase reporting needs to match the store. Two practices carry most of the weight:

1. Send events from the server as well as the browser, with proper deduplication so a single purchase is counted once. The reasoning and the build are in server-side tagging.
2. Include content ids, content type, value and currency on every commerce event, using the same id convention as the catalog.

Expect the platform’s purchase count to differ from the store’s order count even when everything is configured correctly, because the two systems count different things over different windows. The structural reasons are in why does Meta report more purchases than my store, and the diagnostic order when the gap is large is in Meta purchases not matching orders.

Campaign shapes worth running

Three configurations cover most catalogs:

• Retargeting from site activity. Audiences built from viewed, added to cart and purchased events, with exclusions so recent buyers are not shown the item they just bought. Set the exclusion window from your replenishment cycle rather than a default.
• Broad catalog prospecting. The catalog shown to a broad audience with the delivery system choosing items. This works when the catalog is large, the data is clean, and the product sets carry the economics.
• Cross-sell and upsell. Complementary sets shown to recent purchasers, which fits categories with genuine accessory or refill relationships.

Which objective each of these should use, and how they interact with automated shopping campaign types, is covered in which Meta campaign objective should an online store use and should stores use Advantage Plus shopping campaigns.

Creative on top of the catalog

Catalog ads do not have to look like a product grid. Frames, overlays and intro cards can wrap catalog items in brand styling, and collection formats pair a video or image with catalog items underneath. Two constraints: any overlay text must stay clear of the crop boundaries used by different placements, and every claim in the wrapper is subject to the same advertising standards as a standalone ad.

The QA pass before launch

Run this list before a campaign spends anything:

1. Catalog item count matches the number of sellable products, within the tolerance you expect.
2. No items flagged with errors that prevent them being shown.
3. Three sample products checked end to end: catalog id, event id, landing page price, availability.
4. Product sets return the expected number of items, and a newly created product lands in the right set.
5. Exclusion set applied to every campaign, including restricted and out-of-scope items.
6. Purchase deduplication verified with a test order.
7. Prices and sale prices match the site, including tax display rules for your market.

Maintenance

Set a weekly check on catalog errors and item counts, a monthly check on product set membership after any merchandising change, and an immediate check after a theme release, an app change or a catalog import. Feed quality is the ongoing work, and it sits inside product feed optimization alongside the shopping surfaces that read the same data.

---

Sources

• Meta: Advertising Standards
• Google Merchant Center Help: Product data specification