Facebook Andromeda Update

imccafey

Elite Member

Jr. VIP

Aug 24, 2020

5,128

1,662

• Dec 2, 2025

• #41

Going by what others have said, cutting obvious losers is fine, but spamming dupes isn't. Some massively diminishing returns on that. Cull weak ads, keep your best performer fairly as-is, and feed the system fresh creatives when performance dips. New signals will still be preferred over duplicated signals.
I wonder how OPs going with this now though.

 

varrper

BANNED

Sep 8, 2025

282

30

• Dec 5, 2025

• #42

Just turn them off and add new ones. I do not recommend deleting them, because optimization and training may go astray.

 

starpareek

Junior Member

Sep 30, 2011

109

13

• Dec 13, 2025

• #43

wfjcc said:

I’m running ads under the new Andromeda update with a CBO setup—1 campaign, 1 ad set(Advantage +), and 5–10 ads. When some ads start underperforming, is it better to cut them and add new creatives, or just remove the weak ones and duplicate the top-performing ads instead?

Click to expand...

From what I’ve seen, don’t just duplicate winners over and over — that burns them fast. I usually cut the weak ads, keep the winners running, and add fresh creatives regularly. That keeps CBO learning healthy and results more stable.

 

Rehmaann707

Junior Member

Jr. VIP

Nov 30, 2025

161

8

• Dec 13, 2025

• #44

To maintain your campaign more steady, it's typically preferable to replicate the top-performing advertisements and swap out the poor ones.

 

B

BulkCloudStore

BANNED

Mar 23, 2024

125

23

• Dec 13, 2025

• #45

Do not replicate victories under Andromeda + CBO; instead, remove the subpar advertisements and replace them with new creatives, allowing the system to continue learning from the first winners without resetting performance.

 

T

tade818

Newbie

Dec 13, 2025

11

6

• Dec 13, 2025

• #46

I'm currently testing a setup that is working like a charm with a $30 daily budget
Total Ads: 9 | Ad Sets: 3 (with 3 ads per set). | Inside each set, i run 2 retargeting ads and 1 prospecting ad.
It’s performing really well for me right now, though obviously, it depends on the region

 

mylove_14k

Junior Member

Jr. VIP

Mar 27, 2025

110

8

• Jan 21, 2026

• #47

Turn off weak ones and add new ones with potential based on data from the old ones. Don't overwrite each other within the same group to avoid messing up the algorithm.