Creative Diversification vs Doubling Down on Winners: Advanced Facebook Ad Creative Testing

Quick answer

Allocate 70% of spend to your top three winners, 20% to the next three challengers, and 10% to fresh tests. This 70-20-10 split lets you compound the winners while still discovering the next ones. Single-creative accounts always die; over-diversified accounts never reach scale.

Why both extremes lose

Doubling down on one creative means you have a single point of failure. When it fatigues, your account collapses overnight. Spreading budget evenly across 20 ads means none of them gets enough volume to exit the learning phase. The winner-loser distribution is power-law — a few ads carry the account — so the right answer is to bet big on the proven winners and use the rest as exploration capital.

The framework, step by step

Rank every active ad by 7-day CPA and 7-day spend.
Allocate spend in three buckets: top 3 winners (70%), next 3 challengers (20%), all tests combined (10%).
Inside CBO, this is enforced by ad set structure: one evergreen ad set with the top 3, one challenger ad set with the next 3, one testing ad set with new creatives.
Move ads between buckets weekly based on CPA performance.
Promote tests to challengers when they hit account median CPA. Promote challengers to evergreen when they beat the bottom of the top 3.
Demote evergreen ads when they fall below the top 3.

Example test matrix

Bucket	% of spend	# of ads	Promotion rule
Evergreen winners	70%	3	Beats challenger CPA
Challengers	20%	3	Beats account median CPA
Tests	10%	5-15	Beats £50 threshold rule

Pitfalls to avoid

Diversifying too widely. 20 active ads at 5% each starves all of them.
Doubling down too narrowly. One ad cannot carry £20k/month spend forever.
Mixing buckets in a single ad set. Use separate ad sets so the auction respects the allocation.
Promoting ads on CTR alone. CTR is necessary but not sufficient.
Forgetting to refresh the testing pool. Stale tests never produce challengers.

5 FAQs

Why 70-20-10? It is robust across spend levels from £3k/month to £100k/month. Tighter allocations break.

Can I run only an evergreen bucket? Only at the cost of long-term scale. Without the test bucket, you cannot replace fatigued winners.

What if I have only one winner? Drop the challenger bucket and use 80-20 (evergreen-test) until you produce a second winner.

Should the buckets be separate campaigns? Separate ad sets inside one CBO is fine. Separate campaigns is cleaner reporting.

How fast should I move ads between buckets? Weekly. Faster than that creates noise.

Build the bench

Pix-Vu produces enough variant volume to keep the testing bucket full every week, so the winners pipeline never runs dry. https://pix-vu.com.