This message was deleted.

I think I might know what's going on there. The uplift graph is lognormal and the bigger the value, the more skewed it gets. I have a feeling the current graph bounds calculation is not taking that into effect. I'll take a look

congrats on that test!

+1 to that! this is also a problem when there are couple of metrics on this chart that have close to zero lift, and then one of them has something like +50% ..+100% an edge case, but an important one