Ok. Here's the simple explanation.
Conventional Path: The data is inserted in the conventional manner. Each insert is processed separately. The database looks for free space in blocks and uses it if possible.
Direct Path: The data is built into new blocks and loaded after the table high-water-mark. There is no attempt to fill up unused space in existing blocks. Bypassing all this extra work makes it much faster.