I need to measure and communicate to non-statisticians the usage pattern of a software component. It seems that based on data the usage follows a nice smooth heavy-tail distribution: lots of users don't use the component at all, few users use it a lot.
I don't want to summarise my report using a single mean, since I believe it hides the true nature of the data and non-experts will easily interpret it as it was a normal distribution. On the other hand I can't plot the distribution and start explaining what is power-law, why it doesn't look like normal distribution, why "outliers" are not really outliers but well expected users etc, etc...
So I ended up splitting the distribution in 3 parts: users that make some small use of the component, average users, and power users (that use component a lot). For each group I measured mean values (and error of the mean) and in my opinion reporting these few numbers is something more actionable and easier to digest by non-experts.
I wonder if there is some better way to perform that kind of analysis for the non experts. Any suggestions more than welcome!