When catalog performance moves, do not start by mechanically increasing budget. First verify that Meta receives consistent product and event data and that campaign structure reflects your economics.
1. Product-data consistency
Titles, categories, prices, currency, availability, links, and identifiers should mean the same thing in the feed, on the site, and in events. Review errors and warnings in Commerce Manager and fix the root cause in the source system, not only in the exported file.
2. Eligibility and business rules
Ads may include products that are technically active but poor candidates for promotion: unavailable in the target region, incomplete, slow to ship, or unable to support the acquisition cost at their current margin.
3. Pixel and Conversions API consistency
Browser signals can be incomplete, but server-side delivery is not a reason to duplicate events or bypass user consent. Compare event names, time, value, currency, and item IDs. For events sent through both channels, configure deduplication with a consistent event ID.
4. Audiences as suggestions, not prerequisites
In Meta Advantage+, audience inputs can act as suggestions before the system searches more broadly. A lookalike or customer list may be a useful signal when the data is sufficient and lawfully collected, but it is not a mandatory launch condition.
5. Creative and delivery context
Catalog format does not remove the need to explain why someone should buy from you. Review images, titles, price, offer, and landing page. Separate acquisition and re-engagement messages only when those segments can be measured reliably.
6. Diagnostics in Commerce Manager
Use Commerce Manager—not Google Merchant Center—to diagnose a Meta catalog. Review rejected items, data-source issues, event-to-item matching, and product availability in the markets you target.
7. Campaign structure based on economics
One campaign may be appropriate when products have similar economics and enough data for optimization. Separation can make sense when budgets, countries, goals, margins, promotions, or availability rules genuinely differ.
Conclusion
This checklist is not a forecast. It helps separate data, setup, and business-logic issues from normal auction variability. Record the baseline, change one material element at a time, and assess results within an agreed attribution window.
