Feature Focus
Product Stop Loss / Exclusion
Overview: Isolating Performance Lift
BigAtom’s Product Exclusion feature automatically identifies and removes underperforming products within a specific campaign context or at account level. This ensures that ad spend is directed only toward the SKUs most likely to convert for the selected audience — while still allowing those same products to perform in other campaigns where they are strong.
To validate the real incremental gains in ROAS and campaign efficiency, we use a controlled and measurable A/B Testing Methodology.
I. The A/B Test Methodology — Head-to-Head Comparison
The experiment compares performance between two identical campaigns — one running on an Optimized Product Feed (with Product Exclusion active) and another on a Standard Catalog Feed (with no exclusions applied).
1. Test Setup
Element | Campaign A (Treatment) | Campaign B (Control / Holdout) |
Product Feed | Optimized Feed (Product Exclusion Active) | Standard Feed (No Optimization Applied) |
Optimization Logic | Poor products removed only based on Group A performance | Full catalog always included |
Campaign Structure | A parallel set of similar campaigns | Identical parallel set |
Goal | Measure incremental efficiency and ROAS lift | Establish baseline performance |
2. Campaign Setup Requirements
Control variables ensuring clean and valid data. The following must match between Group A and Group B:
Audience Targeting — Lookalikes, custom segments, or broad targeting must be consistent
Budget Allocation — Daily and total budget must be equal
Bidding Strategy — e.g., Lowest Cost, Value Optimization must match
Campaign Objective — All campaigns must use the same goal (e.g., Sales / Conversions)
Creative Assets — Ideally the same dynamic creative templates or similar formats
Any deviation in these variables may bias the results or invalidate the comparison.
3. Test Duration & Data Requirements
Requirement | Guideline | Rationale |
Minimum Run Duration | 4 Weeks (28 Days) | Allows sufficient delivery cycles + post-learning stability |
Minimum Conversion Volume | 100+ conversions per group | Ensures statistically reliable ROAS & CVR measurement |
Daily Budget | ≥ (Account Avg. CPA × 50) ÷ 7 | Supports fast learning exit and consistent delivery |
Product Count | Only 20%–30% of total catalog and ≥ 1,000 products | Prevents spend scattering and ensures Meta has enough variety to optimize |
“During Week 1 or until the campaign achieves 50 conversions (whichever is earlier), both Group A and Group B must run without exclusions. Once Meta exits learning or crosses the 50-conversion threshold, the Product Exclusion logic will be activated in Group A only.”
II. Success Metrics — Measuring the “Goodness”
1. Primary Performance Indicator
Metric | Expected Outcome | Reasoning |
Return on Ad Spend (ROAS) | ROASA > ROASB | Direct evidence of increased revenue efficiency due to product exclusion |
2. Secondary Efficiency Indicators
Metric | Expected Outcome | What It Proves |
Cost per Purchase (CPP / CPA) | CPPA < CPPB | More efficient customer acquisition |
Conversion Rate (CVR) | CVRA > CVRB | Stronger relevance and purchase intent |
Conclusion
This A/B experiment method ensures a scientifically fair evaluation of how Product Exclusion improves Meta campaign performance. It isolates the true incremental gain — proving that removing low-performing products results in:
Higher ROAS
Better budget efficiency
Increased likelihood of conversion
Reduced wasted ad spend