Browsable folders

Claim: a collection is shallow and broad: two levels rather than three, and folders a reader can scan, of at least 3 and at most about 32 entries.

What it asks

  • Prefer breadth to depth: a third level is the costly one.
  • A folder of 1 or 2 entries folds into its parent, unless it is one record of a collection (a job, a chapter, a topic).
  • Past about 32 entries, group by purpose. A folder reached through an index file, and read rather than browsed, may hold more.
  • The entries' names carry the scent (names predict content).

Why

Two levels of 16 by 32 took 36 s, 32 by 16 took 46 s, and three levels of 8 took 58 s, among three web sites of 512 items each, with 19 participants. The two two-level layouts did not differ significantly, and only the three-level one was slower (Larson 1998). Among 64 menu targets, two levels of eight beat both deeper and flatter menus (Miller 1981). Needless depth costs more than breadth.

 time to find an item among 512 (Larson 1998)
 one mark per 2 s
 16 × 32, two levels      ██████████████████          36 s
 32 × 16, two levels      ███████████████████████     46 s
 8 × 8 × 8, three levels  █████████████████████████████ 58 s

Limits

The studies timed search for a named target through category labels. Developers mostly reach code by search, about twelve queries a day (Sadowski 2015), and spend about a third of their time navigating (Ko 2006). An optimum of 8 to 16 items per level, often quoted, is not what these studies found. The floor of 3 is not measured: a folder of one or two entries adds a level for little, and these studies show that a level costs, but none tested a level so small.