The scheme depends on what the user already knows

What it is

Whether to use letters, time, topic, or task is not decided by what the content “really is.” It is decided by what people already hold when they start looking. That is organization follows known attributes: contacts are alphabetical because people usually remember a surname; a policy hub is task-based because people remember the outcome they need, not a document number. An organizing axis matches a set of knowns. It does not reveal the essence of the content.

Knowns change. New users may know only a task, return users may remember a location, support staff may hold only a ticket number. The same collection should match different knowns for different roles, rather than crowning one “correct” classification for everyone.

Why it happens

Lookup constrains the unknown with the known. If the axis is bound to an unknown attribute, constraint cannot start, and people can only browse or quit. An alphabetic axis requires known spelling; a time axis requires known when; a topic axis requires a word from the right domain. The wrong axis demands a translation step: turn “I need to file expenses” into the architect’s “Finance / Policy / circular 2023-07.” That translation is extra cognition. When it fails the taxonomy looks bad; the root is a mismatch between axis and knowns.

What users know cannot be imagined. It has to be taken from the clues still available at the moment of the task. Retrospective interviews backfill attributes that were not present then (“of course I knew it was Finance”); those attributes must not choose the axis.

Studying it

Measure the clue distribution first, then choose axes from it—do not freeze a classification and hunt for justifications afterwards.

• Paradigms: critical-incident interviews and query logs that record the attributes actually spoken or typed at launch; then contrast candidate axes against those attributes.
• Independent variables: which known attributes are given to participants, and whether the organizing axis matches or mismatches them.
• Dependent variables: frequency distribution of launch clues, the success gap between matched and mismatched axes, and whether a translation step appears (people rewrite their own clue to fit a category name).
• Methodological note: card sorts answer “do these things seem alike,” not “which fact did you know when you looked.” Choosing an axis from classification agreement yields a scheme that is kind to cataloguers and cold to lookup. New and returning users often have different knowns; stratify, do not average.

Where it stops holding

Exploratory browsing has no stable known, so the axis cannot be chosen from “known attributes”; it has to be chosen from scannable topical differences. Knowns also grow mid-task: after finding one piece, people may acquire a date or a number, and they should be allowed to switch axes rather than being locked to the launch kind. Forced registration and permission walls strip some knowns (“my applications” is invisible when logged out); an axis that holds for anonymous users may not remain the best match after login.

Applying it

• For each critical lookup, write the launch-time knowns: words, times, places, roles, task names. Pick axes only from high-frequency knowns.
• When several knowns are frequent and do not translate into each other, provide several axes rather than voting for one “primary” classification.
• Category names must catch that known. Do not translate “expense claim” into an internal circular number.
• Verify by stopping someone about to look and asking only “what do you remember about it right now.” Check the answer against the first navigation layer. A mismatch means the layer was built from the architect’s knowns, not the user’s.

Related

• Within the group: G1.02.1 Exact schemes (alphabetical, chronological, geographical) have one right place · G1.02.2 Ambiguous schemes (topic, task, audience) depend on judgment · G1.02.3 The same content can live in multiple schemes
• Adjacent: G3.01 Searching versus browsing · G2.08 Predictability of navigation items · Q2.11 Card sorting
• Search terms: known-item search · organization schemes · information scent