Comment by daguar

This is neat! That said, I don't think the statistical size makes this a conclusion you can draw:

- "a reply rate of 1% (1 in 100 users)"

- "a reply rate of 4% (1 in 25 users)"

- "a reply rate of 8% (1 in 12 users)"

Sorry to say but... we really can't draw a conclusion like this strictly from these kinds of numbers.

Totally possible these conclusions ARE true, but I'd test at a bigger scale and/or with qualitative research with users.