I used this once and was really happy with the results. I think there's a little bit of an opportunity bias which good statisticians would have issues with, but for some useful imperfect numbers I think it's a good solution.

Also, this is a pretty new tool.