I'm not sure what the n.0.0 part is but the very fact that the x_pct computations share a common denominator seem to indicate that a specific point was missed...
Also, and more importantly, there is no breakdown by msg which is the important thing being counted.
Each msg count needs to be measured by the particular appVersion, then those computations compared against each other.
This is to see for example, that some message's frequency wrt to one appVersion, has increased/decreased compared to another appVersion. In this case the denominator is the total number of msgs for that particular appVersion, not a global total.
I added a comment with an example (I didn't see a way to edit my question) but apparently the comment hasn't been cleared by the moderator yet.
Sorry for any confusion. Please wait for the example to clear if the context is not clear.
... View more