-
I tested Meta’s Andromeda creative refresh thresholds across 47 ad accounts — here’s the data on when fatigue actually kicks in
Posted by on May 3, 2026 at 8:52 amSpent the last 90 days pulling fatigue data across client accounts ranging from $5K to $180K monthly spend. Wanted to share what I found because most of the "refresh every 14 days" advice floating around isn't matching what I'm seeing post-Andromeda.
Setup:
- 47 ad accounts, mixed verticals (DTC beauty, SaaS, fitness, fashion)
- Tracked hook rate, CTR, CPM, and frequency from launch through fatigue
- Defined fatigue as the point where 3-day rolling CPA exceeded launch CPA by 25%
What I found:
The 14-day rule is dead. Median time-to-fatigue is now 8.3 days for static creatives and 11.7 days for video. Andromeda is identifying creative repetition way faster than the old system.
Hook rate is the leading indicator, not CTR. CTR drops happen 3-5 days after hook rate already collapsed. If you're waiting for CTR to tell you something is wrong, you've already burned $1,500-$3,000 in suboptimal delivery on a typical ad set.
Frequency caps are misleading now. I had ad sets with frequency of 1.4 already showing fatigue signals. Andromeda doesn't care if your audience has "seen" the ad — it cares if the creative pattern matches something that already underperformed across the network.
The "edit existing ad" trick still resets learning phase about 60% of the time. Duplicating with new creative inside the same ad set has worked better in my testing — keeps the audience signal intact.
What I'm still figuring out:
- Whether creative diversity at the ad set level vs campaign level matters more
- If there's a way to avoid the learning phase reset entirely (the "edit-in-place" approach Meta mentioned in their docs is inconsistent in practice)
Curious what others are seeing. Anyone tracking hook rate as their primary signal yet, or still relying on CTR/CPA?
-
1 Reply
-
how much spend per ad until you shut it off
Log in to reply.