My understanding is that Direct Discovery is meant to let you reference extremely large data sets (Big Data) without having to load the extremely large data set into memory. You can only reference ONE data set via Direct Discovery in the current release.
In a data warehouse using a star schema or something similar, you will find dimension tables and fact tables. Fact tables are typically larger and grow rapidly compared to dimension tables. An example fact table is a revenue table. An example of a dimension table is a time dimension table (revenue periods, etc), product category dimension table and a customer table. Assuming your fact table contains references to the dimension tables, it is possible to see the revenue information BY customer or BY product category or BY time period or by a combination of customer and category and time dimension.
In this example you would want to load the time, customer, product category data directly into QlikView and have it in-memory. The revenue fact table (if extremely large) would be where direct discovery would come into play. In this way you would be able to aggregate revenue based on data that is in-memory.
It really depends on what you are trying to achieve...you may not need direct discovery at all if it's not that big or if you have a large server with many GBs of RAM. Also, depending on how your data is stored in the DB and what you need to do you may need to load parts of the data in-memory and reference another part via direct discovery.
I am still learning about direct discovery...so this is just my very basic understanding. Hope it helps.