What I like about Graylog is that it's real-time and you have access to the raw data. So, you ingest it, and you have access to every message and every data item you ingest. You can then build analytics on top of that. You can look at the raw data, and you can do some volumetric estimations, such as how big traffic you have, how many messages of data of a type you have, etc.
