Furniture Dynamic Product Ads: Catalog, Creative & Tracking Checklist

Separate the platforms before changing the feed

For Meta, review the catalog connected to the campaign, the selected product set, and the event integration used by the store. Keep a record of what the integration sends and what the catalog contains. Avoid changing IDs simply to make them look like your internal SKU labels.

Google’s retail dynamic remarketing uses a Merchant Center product feed and a linked Google Ads account. Its retail setup guidance describes that connection. Shopping ads and dynamic remarketing can use product data, but they are different advertising uses; a valid feed does not prove every campaign or event configuration is ready.

Write down the account, catalog or Merchant Center source, feed owner, integration, currency, target market, and campaign being reviewed. If another team owns stock or pricing, agree how errors reach that team and how you will confirm the fix.

Run five checks on each test product

Choose products that expose different conditions: a simple item, a finish or size variant, a discounted item, and an unavailable item. These are suggested test cases, not a required sample size or evidence that the rest of the catalog is correct.

Keep each failing item identifiable in the worksheet. Fix the source of a repeated error where possible, rather than manually patching one exported row that the next sync will overwrite. Recheck the processed record and the destination after the next update.

Keep variant IDs and event mappings explicit

Google’s item-group guidance distinguishes an individual product’s unique ID from the shared group ID used for variants. Submit each variant with its own record and appropriate distinguishing attributes. Use the actual manufacturer identifiers where applicable; Google’s identifier guidance says not to make up a GTIN.

For Google retail dynamic remarketing, the event-parameter documentation says the item id must match the feed’s id, item_group_id, or display_ads_id. Verify which mapping your integration uses. Do not assume the Google Ads remarketing payload is identical to a separate analytics event.

Meta’s Business SDK parameter reference distinguishes content_type values for individual products and product groups. Check that the identifiers and declared type describe the same level of the catalog. Changing all group IDs to variant IDs without checking the integration can create a different mismatch.

Record what the browser and any server integration actually send for each tested action. If both report purchases, have the implementation owner verify deduplication in the platform’s diagnostics. Confirm value and currency against the transaction definition. Keep test activity separate from business results and respect the store’s consent configuration during testing.

Use AI images only when the product and destination rules hold

Start with a clear view of the exact item for sale. Compare generated outputs with the source photo for finish, legs, seams, cushions, handles, proportions, and orientation. Reject an image that changes the product or suggests included pieces that are not sold with it. A generated room cannot establish dimensions, material performance, or assembly behavior.

Google’s image-link requirements prohibit promotional overlays and require an image of the actual product. The same guidance requires generative-AI images to contain metadata identifying their AI origin and says to preserve embedded source metadata. Check the final exported file, including any resizing or conversion step; an original generation file does not prove the uploaded version retains it.

A clean product image and a room-context image answer different questions. Compare eligible alternatives under a defined test; do not assume that a lifestyle scene always beats a product cutout. Check requirements for the specific placement before reusing an asset across platforms. The AI-tool review checklist covers output acceptance, and the content workflow covers approval and version control.

Follow the click all the way to the selected product

Google’s landing-page requirements address consistency between submitted product data and the destination. Inspect the page a new shopper actually reaches, including any location or currency selection. A catalog URL that loads successfully can still open the wrong finish or display an outdated offer.

Also check a sold-out variant and an expired promotion. Identify whether the feed, processed catalog, storefront cache, or destination selection caused the discrepancy. A fix is complete when the relevant shopper experience and platform diagnostics agree.

Define a campaign test that you can interpret

Once the catalog and events pass review, write a test brief: eligible products, audience, creative version, destination, spend limit, outcome definition, and decision date. Choose audience windows from the campaign’s available settings and your observed buying journey. A fixed sequence of retargeting windows is not a universal furniture standard.

Separate prospecting from retargeting and existing-customer campaigns in your reporting. Record attribution windows and whether revenue includes modeled or view-through conversions. Check actual orders, discounts, cancellations, returns, and contribution after delivery and advertising costs. Revenue divided by ad spend is ROAS; it is not profit.

Do not add up revenue credited by several platforms and call it total sales. A sale can appear in more than one report. Nor does a high retargeting ROAS establish how many additional sales the ads caused. A holdout needs a suitable design, sufficient observations, and controls for other differences; an arbitrary percentage and duration do not guarantee an incrementality estimate.

If there are too few verified outcomes, report the limit and the next observation needed. Use the Meta campaign worksheet or the Google Ads guide for campaign planning, and the measurement guide to connect platform reporting with business records.