http://snowplowanalytics.com/ -- this is an another Open Source project which you can use in house and a loads and loads of predefined metrics and documentation.. but I agree the approach here is simple and rock solid, plus you have complete control of it. Just wondering if JSOB datatype is used to store the unstructured data into Postgres?

Snowplow looks very promising but we ruled it out because of the dependency on AWS. And we use the hstore data-type.

I did the similar thing but with nodejs to collect and process the events and Postgresql JSONB for storing the events .... i see the advantage on this approach is that i dont need to do much for structuring the data in node as i am just dumping this unstructured data directly into JSONB column , but i believe in your case you are structuring it first in Django and then storing it in hstore .. I totally agree with you about the Visualisation .. thats the hard part .. in my case i have totally unstructured events but still i managed to use a simple node app to give me the data back by querying JSONB .. was it easy to query HSTORE??