Thread: Performance Max Feed Testing Strategy

1/ Performance Max is a black box, but your feed isn't. Here's how we test multiple feed variations without screwing up your campaigns:

2/ The problem: PMax campaigns optimize toward your existing feed. If you change titles mid-campaign, you lose learning. If you don't change them, you can't improve.

3/ Solution: Feed supplementation, not replacement.

Create supplemental feeds that override specific attributes for specific products. Your base feed stays intact.

4/ The testing structure:

• Control: Original feed (no changes)
• Test A: Benefit-focused titles
• Test B: Spec-focused titles
• Test C: Keyword-optimized titles

5/ Key rule: Don't test all products at once.

Start with your top 20% of products by revenue. These have enough volume to reach statistical significance faster.

6/ How to measure:

• Export product-level performance weekly
• Compare click-through rates by feed version
• Track impression share changes
• Monitor conversion rates (not just clicks)

7/ Timeline expectations:

• Week 1-2: Learning phase, don't judge yet
• Week 3-4: Patterns emerge
• Week 5+: Statistical confidence

8/ Common mistake: Testing too many variations.

Three feed versions is usually enough. More than that, and you're diluting your budget across too many experiments.

9/ What we've seen work:

• Adding brand name to titles (when missing)
• Front-loading benefits over specs
• Including price points in descriptions
• Adding "free shipping" when applicable

10/ What usually doesn't move the needle:

• Minor word order changes
• Synonym swaps
• Adding more attributes that Google ignores

11/ Final tip: Document everything.

Screenshot your feed rules, export baseline metrics, and track every change. Future you will thank present you.

12/ Questions? Reply with your PMax feed challenges. Happy to riff.