Continuing the discussion from Measuring OpenMRS Impact:
I would like to separate the “what” from the “how”, or else @janflowers original thread is going to become a huge mess.
On the topic of “how”, I agree that instrumentation of OpenMRS installations is a valuable piece of this, but also that implementations and networks may not be so thrilled to share this with OpenMRS without some real reason.
I think OpenMRS Atlas is where the OpenMRS Community should store aggregated statistics about the usage of OpenMRS in different sites. (If we don’t think that’s the case, we might as well throw Atlas away.)
The next Reference Application release is going to include a nag to the administrator to configure it, so hopefully we’ll see an uptick in the amount of reliable and timely data that goes into it.
We could make sure it has an easy interface so that different tools can post updates. E.g. what @hamish mentions in Rwanda, what @janflowers mentions in Mozambique, and whatever derives from @lober’s comments should all be able to submit data. We don’t need to be prescriptive about that, and the technical piece of this is pretty trivial compared to the political buy-in neede to submit some aggregate data to OpenMRS.
(A crazy thought: we should use DHIS2 as the back end for this. 
It seems like an appropriate use of that technology, and it would be nice of us to leverage the work of an open-source colleague where we can.)