Significantly higher "baseline" CVR in PPO A/B test than observed in "Metrics" section

Hello!

I am running a PPO test where the "baseline" CVR is more than twice as high (+5pp) as the CVR I see when I look at the overall CVR in the "Metrics" section of the Analytics tab. I know PPO CVR is "estimated CVR" but it seems like a poor estimate if it is >2x the actual observed CVR.

What are the methodological differences between PPO test CVR and actual observed CVR? What could explain such a wide variation in CVR?

Significantly higher "baseline" CVR in PPO A/B test than observed in "Metrics" section
 
 
Q