1. Avoid Over-Restricting Product Volume
Do not significantly limit the number of products in a Product Set if the account is historically optimizing across a much larger catalog.
Meta’s algorithm performs better when it has more options to optimize from.
Recommendation:
Populate at least 200+ products or 30% of your total catalog, whichever is higher.
2. Use Meta-Attributed Metrics for Filtering (When ROAS Is the North Star)
If ROAS is the primary objective, rely on Meta-attributed metrics for inclusion or exclusion logic rather than website/backend numbers.
This ensures:
Alignment with how Meta optimizes delivery
Better consistency between optimization signals and reporting
3. Always Pair Metrics with a Spend Threshold
When setting inclusion or exclusion rules, avoid filtering purely based on performance metrics (e.g., ROAS, CPA, Conversion Rate).
Instead, combine performance conditions with a minimum spend threshold to ensure sufficient data before taking action.
How to calculate spend threshold:
Let a product spend at least 1x the current Cost Per Sale (CPS) at account or campaign level
In certain cases, allow up to 2x CPS before exclusion
This prevents premature removal of products without statistically meaningful data.
4. Adapt to Meta’s AI-Driven Optimization
With Meta’s increasing reliance on AI and automation:
Provide more product options to the system
Control efficiency using product-set-level exclusions, rather than overly restrictive inclusion
This approach:
Allows Meta to explore and identify winners
Forces deprioritization of underperforming products
Improves alignment between product performance and audience targeting