>Identify the purpose of a screenshot
The author emphasizes this as one of the important considerations.
Definitely a good idea but I think people need to go further.
I would say identify the purpose of the document to start with.
Ideally, I would like to have a complete reference tome where the text could actually stand on its own extremely well if it had to, along with detailed essential illustrations such as charts or schematics. Any other drawings or screenshots should serve as the icing on the cake without actually being essential to understand the documentation.
If lots of illustrations are intended to be mixed with the final text anyway, this type of document can still ideally be drafted and revised in text only format, quickly leaving placeholders for future illustrations as you go along. All you need at this point is a pretty good idea of what kind of picture you would put there after you are done. You'll also have more time later to think about the exact ilustration you'll want for each placeholder, and better able to select or create more meaningful pictures the first time. Once the final draft of the text is complete, then (only) the most helpful illustrations can be introduced to their expected positions. Then without (much) further editing to the text, consider the initial placement of illustrations to be a first draft of artwork, then it may only be necessary to manipulate the artwork alone a little bit more before a final draft of that complete document is reached.
I think reference documentation should be complete and text-heavy.
If there is actually no real reference documentation, then whatever is documented will need to be more text-heavy to compensate.
OTOH the the heavily-illustrated approach should be reserved for more specific topics such as detailed tutorials or quick-start guides.
Different documents having different purposes, compiled into a full documentaton package.