Grouping reduces cost more reliably than cutting items
Aliases: chunking navigation · do not cut first
What it is
Faced with too many navigation items, the first impulse is often to delete a few. Grouping usually cuts cost more than cutting: split one choice field into named chunks, and both scanning and deciding happen inside a chunk. Total cost drops while the entries remain. Cutting reduces coverage. Grouping changes the shape of the choice field. Group first when grouping is possible; deletion is the last knife.
Why it happens
n items in a flat row are one choice. After k labelled groups, people choose a group and then an item inside it: two small choices instead of one large one. Hick-style cost becomes roughly “choose group” plus “choose inside”, usually less than one choice among n. Scanning becomes read the group titles, then only the relevant block; the eyes need not cover the whole column. Coverage is not lost: entries still exist, they are just not in the first-glance field.
Cutting removes coverage. If a deleted item is still some people’s main path, they detour or conclude the function is gone; cost moves from this row into search, support, and churn. Only when the item has no real task, or a better arrival already exists, does cutting reduce cost rather than relocate it. Grouping fails when group names are vague (More, Other): people still have to open every group to learn what is inside, two choices become two full scans, and cost rises.
Studying it
Take the same entries and compare flat, meaningfully grouped, and rare items deleted. Measure time to frequent and rare targets, errors, and judgements of “does this still exist”. Independent variable: whether group labels are specific. Dependent variables: sum of the two choice times, discovery of rare targets.
If after grouping, rare discovery stays close to flat while frequent arrival speeds up, grouping kept coverage and speed. In the deletion condition, rare discovery should drop; that is the coverage price of cutting, visible by contrast.
Where it stops holding
Too many groups and choosing a group becomes a large choice again; cluster further or offer search. When boundaries fight the user’s model (splitting Bills from Invoices), people enter the wrong group and grouping becomes a penalty. When there is no room to write group names, the grouping cue vanishes and cost returns to flat. Experts who had memorised a flat layout pay a temporary tax when grouping appears; the new places need to stay still for a while.
Applying it
- Name groups in the categories people actually say, such that a name excludes the other groups. More is not a group name.
- Frequent items may also stay on an ungrouped short list; do not hide them in a group for the sake of tidiness.
- Delete only when task logs show no one completing a main path through that item; look for an equivalent entry first.
- How to check: arrival at frequent items should be faster or at least unchanged; arrival at grouped rare items should still complete in two choices. If the group named Other absorbs most failed find-tasks, rewrite the name rather than keep deleting.