-
Why is performance not the same after applying an successful experiment
I was running a 50/50 experiment as a way to grow overall performance:
original: Target CPA
experiment: Manual CPA
The experiment ended getting 3x conversions at lower CPA and a higher impression share, so I decided to apply changes over the original.
After 3 days I feel my campaign has sunk, its like the changes were applied but the performance is nowhere near.
Has anyone experienced something similar?
-
1 Reply
-
Some possible causes on top of my head: Sample size; user not randomized between variations ; sequential comparison so some external factor changed since the rollout.
Log in to reply.