Preceding Load allows you to perform additional load steps before initializing a table. Each step uses the output of the preceding load statement as input.

Example:

preceding load

Source

In my opinion, the biggest benefit of using this is code readability. There may be times where putting all the transformations in a single Load statement is faster, however it is in most cases easier to understand code that pieces out the transformations into multiple steps versus having a ton of nested functions.

With that said, if performance is a high priority it is worth testing various avenues. All in one load, preceding load, exporting to QVDs and performing the transformation on a new ingestion, etc… There will almost always be multiple ways to solve a problem. A priority in any type of development is that it is readable, whether it is you 6 months in the future or a colleague who takes over support.