Many guides I've seen quote a number called CMGR. An example is here:

http://a16z.com/2015/08/21/16-metrics/

CMGR is apparently a bit of a standard, but it provides a number that's in some cases significantly smaller than what you get if you do a straightforward percent growth calculation.

I understand the idea of dampening out variation but what if the variation is small? Why should the results be so hugely different? We're talking 2X here.

I guess I'm looking for someone to go into the math behind the difference. Been a while since I've studied probability and stats, and the origin of the CMGR formula seems murky vs. other ways of computing this. I'm wondering if the CMGR formula carries assumptions that are not explained in any of the text I've found on it.