Rather than spend hours coming up with a complex numbering scheme, this might be an excuse to implement something far more straightforward discovered by an extensive readability study at IBM, of which I was a part. My work involved sitting behind a one-way mirror with a stopwatch, watching people take tests that involved, among other things, "how fast can you find Figure 3-4?" We had cameras mounted over the participant's shoulders and could watch them thumb through the documents, and we also monitored eye movements. Then we followed up with a short interview where we got feedback.
It's been a tradition for years that documents use certain types of numbering systems: specifically, chapter-page numbers, and separate designations for tables, figures, etc. We found that readers were confused by the presence of both tables and figures, and even more confused by individual sequences. For example, participants found that the system of
was confusing, but the individual sequences (items are numbered among themselves), as in
was even worse. IBM's solution was to call everything a figure, even if it's really a table, flowchart, plate, or whatever, and to number them sequentially. And it's a heckuva lot easier to implement, especially if you're stuck using Word.
Finally, readers strongly opposed chapter-page numbering. They reported that the time to find information was greatly increased because as they flipped through the pages, they first had to find the chapter and then the page. Opposition was stronger with shorter documents where a chapter was under roughly 50 pages.
The only evidence I have left are several entries in the book entitled IBM Information Development Guideline: Style. In that book there are several areas where it's mentioned, but this one stands out (I've shortened it a bit for brevity):
In IBM manuals, no distinction is made among different kinds of illustrations. All are called figures, including drawings, tables, plates. . . . All are . . . numbered consecutively within a chapter.
IBM has stopped making distinctions . . . because customers want consecutive numbering of illustrations. . . . The consecutive numbering, they report, helps them find figures more quickly. It is permissible to characterize a figure by using descriptive wording at the beginning of a caption:
Figure 2-3. Diagram of...
Figure 6-4. Table of...
Figure 8-2. Flowchart of...
You notice that the examples use chapter-page numbering. The style guide I refer to here did not cover the overall layout of books from that standpoint; that was addressed in another guide, but I don't have a copy and therefore I can't quote it. But I can tell you that it asked that chapter-page numbering be reserved for those 20-pound systems documents.
Chapter TOCs. In general they were not used very much. Readers were rather ambivalent, saying there wasn't much difference between flipping back to the main TOC or the chapter's TOC. About the only case where chapter TOCs were used were for very large manuals such as the IBM theory of operation manuals. Here, the chapter TOCs went to lower level than the main TOC and readers could find them easily because they were always assembled just after the tabbed pages (in some books they were printed directly on the tabbed pages; readers really liked this).
For books that did not have tabbed chapters, readers didn't use them because they took longer to find than the main TOC, so they were dropped.
For books where the main TOC went to the same level as chapter TOCs and where the chapters were not divided by tabbed inserts, IBM decided to drop them because the labor and extra paper had no benefit and they were so rarely used by readers.
Readers liked the source to take responsibility. Most tech writers would never say something like "we suggest that" and instead would say "it is suggested that". It turns out that readers locked on to the "we" approach, and felt that it made the document more personal, more interesting, and, most importantly, made them feel the source was taking responsibility for what was said.
Don't talk down. Don't include "duh" information. Lots of books, particularly in the area of user guides, spend a lot of time explaining things a six-year old can figure out. Further, computers have been around for a while, so you don't need to explain how to use a mouse (if readers don't know, the book will confuse them anyway). Some examples are:
It's important to note that many readers take offense at this material.
Double- versus single-sided printing. Most (just about all, actually) of IBM's "books" were double sided (duplex). We knew it was cheaper, but we also looked at the usefulness: When books were properly bound (say, ringed binders, stapled, or glued) readers preferred them, especially in situations where the writing described a figure on a facing page (no flipping). The findings only underscored the obvious, but at the time IBM was undergoing a major move to coerce employees to use the duplex features of office copiers.
Page numbering / starting at page 1. Some books start the numbering at the very first sheet of paper; some use roman for the front matter and start arabic at the intro or chapter 1. When asked about the professional appearance, we found correlation between the job level of the reader and the opinion. Again, it was the obvious: As the level of the job increased, readers considered the roman - arabic method more professional; lower-level readers did not care.
Readers like footnotes, to a degree. Most technical writing "experts" decry the use of footnotes. But readers considered them helpful, preferring them over text that interrupts the flow, and tended to report that footnotes gives a book a professional look. The "to a degree" means that once a footnote gets longer than a sentence or two, you might consider weaving the material into the body.
Don't overuse notes. By this "note" I refer to small paragraphs of text that are set off from the running copy by some device like an icon or the word "note". A note should be reserved for important information that can't be woven into the body copy, perhaps because it goes off an a tangent. Lots of notes give readers the feeling that they are not important. To some readers (who are also writers) it shows the writer was too lazy to figure out a way to weave it in to the body.
Running headers and footers. Readers tended to use only the top-most level of heading (usually the chapter title), and even that very rarely. When it was used, it was used as a quick way to find a chapter when quickly flipping through pages. Using any lower-level heading was almost never used, and when it was used at all, only in extremely "thick" chapters. Along the same vein, readers looked to the outside corners of pages for information such as determining what chapter they were reading.
Centered page numbers were considered tacky by some readers.
Lowercase entries in indexes were appreciated by the readers who used the index to determine if a term should be capitalized.
Education and perception. As the level of education rose, so did the appreciation of traditional techniques. For example, at lower educational levels, readers did not notice things like front matter being numbered with romans and the arabic 1 starting with the first page of chapter 1; at higher levels, starting at 1 from the very front was considered tacky and gave the appearance of unprofessional work.