Why topics?

Topics are the basis for high-quality information. They should be short enough to be easily readable, but long enough to make sense on their own.

By organizing content into topics, authors can achieve several goals simultaneously:

Content is readable even when accessed from an index or search, not just when read in sequence as part of a chapter. Since most readers don't read information end-to-end, it's good information design to make sure each unit of information can be read on its own to give just-in-time help.
Content can be organized differently for online and print purposes. Authors can create task flows and concept hierarchies for online orientation, and still have a print-friendly combined hierarchy that helps people who do want an organized reading flow.
Content can be reused in different collections. Since the topic is written to make sense when accessed randomly (as by search), it should also make sense when included as part of different product deliverables, so authors can refactor information as needed, including just the topics that apply for each reuse scenario.

Topics are small enough to provide lots of opportunities for reuse, but large enough to be coherently authored and read. While DITA supports reuse below the topic level, this requires considerably more thought and review, since topics assembled out of smaller chunks often require editing to make them flow properly. By contrast, since topics are already organized around a single subject, authors can organize a set of topics logically and get an acceptable flow between them, since transitions from subject to subject don't need to be as seamless as the explanations within a single subject.