In addition to using "process colors" as described here, I will also:
- group larger processes using colored comment boxes with large titles
- utilize the tool annotations to better describe what's happening in plain English
- name the connector lines when it provides additional clarity (defaults are #1, #2, etc.)
(Tangentially related to documentation)
Finally, for larger workflows where I need to make sure I don't lose data records in the process, I will add a record count data checkpoint at the end of the workflow for each of my inputs. This looks like a "data in" stream where I do a unique count on a certain field and similarly do a "data out" stream in the same manner. This not only helps me when I am building the workflow to ensure I am not missing anything along the way, but it gives other users of my workflow confidence that the workflow is operating as designed.