Hello! I want to ask about CUPED feature. (v2.7) I enabled the CUPED feature for only some metrics in the metric overrides. Looking at the query results, it seems like the covariate data was calculated correctly. However, the variance and p-value are exactly the same before and after applying CUPED. Why might this be? Could there be a configuration issue?

@helpful-application-7107 ?

Can you share any screenshots of how you're comparing before and after applying CUPED? If there's little correlation, in some cases it can appear as though there is no difference.

Thank you for your response. I had asked my question based solely on the visualization, but after extracting the raw data, I found that the values had indeed changed. When I calculated Theta separately, as you mentioned, it was relatively low at around 0.15. Since this metric (ARPU) has always had high variance, I thought CUPED might be effective, but it’s unfortunate that it wasn’t. Is there any other effective variance reduction method available in the growth book that I could try? Thank you.

Capping may be useful for you if your ARPU is skewed with a few high values.

We are already capping very extreme values (at the 0.999 percentile), but perhaps we could adjust that level. Thank you again!