Context-sensitive nav changes must be predictable

What it is

Navigation does not have to look identical on every page to be consistent. Enter docs, and the sidebar becomes a toc; sign in, and “Account” appears; unauthorized items drop out. Those are context-sensitive changes. They have to be predictable: people must be able to state a rule—“pages of this kind change like this”—and the rule must still hold the next time that context appears. If the rule cannot be stated, or pages of the same kind each change differently, the change is random. Spatial memory cannot update; every visit starts with a fresh inventory.

Predictable targets the conditions of change, not the cancellation of change.

Why it happens

People form conditional rules for chrome: “if I am in docs, the left is a toc”; “if I am signed in, there is an avatar at top right.” Once a rule is confirmed, the change itself becomes part of location—“the sidebar became a toc” confirms “I am in docs.” When the rule breaks (some doc pages have a toc, some do not; after sign-in the avatar is sometimes left, sometimes right), people cannot decide whether to update the rule or treat the event as noise, and they trust neither.

Unpredictable sources are often experiment buckets, promo slots, silent permission pruning, and one-off “this page is special” exceptions. Exceptions cannot be learned as rules because they will not be met again. Predictability requires that change bind to a context the user can perceive (section, sign-in state, object type), not to an experiment bucket or an editor’s mood. Local nav replacing by branch is learnable precisely because it binds to the perceptible condition “current branch.”

Studying it

Have people induce a rule, then test whether the rule predicts a new page.

• Paradigms: after browsing a set of pages, predict on an unseen page of the same type: “what will the nav add, drop, and where will it appear.” Compare rule-consistent change with a one-off exception.
• Independent variables: whether change binds to a nameable context, whether change is identical across pages of that kind, presence of a single-page exception.
• Dependent variables: accuracy of the stated rule, hit rate of predictions on the new page, search time after hitting an exception.
• Methodological note: staff can say “that page is a campaign so it is excepted”; users cannot. Hide the exception in the materials and watch whether users invent a wrong rule (“campaign pages drop global nav”) and over-generalize it.

Where it stops holding

On one-shot flows (sign-out confirm, in-payment) nav changes are not expected to recur; predictability is low-value and a clear exit matters more. Accessibility-driven permission pruning must happen, but if “why an item is missing” has no signal, the pruning is still unpredictable; an absence caption or unauthorized state can turn pruning into a perceptible condition. Role switches are perceptible context, so swapping a whole nav set can be learned; swapping sets inside one role with no cue cannot.

Applying it

• Write, for every nav change, a condition the user can perceive (section, sign-in, object type). Changes without a condition get deleted or made permanent.
• Pages of one kind share one change set; ban single-page exceptions. If a campaign page must differ, let it look like a different page type, not like a content page with three items missing.
• Express permission pruning as absence or a locked state; do not let an item exist on some pages and not others with no explanation.
• Verify by showing someone uninvolved three pages of one kind and asking them to state “what the nav does on this kind of page.” Test with a fourth. No rule, or a fourth page that breaks their rule, means the change is still not predictable.

Related

• Within the group: G2.07.1 Navigation position and order must stay stable across pages · G2.07.3 When navigation vanishes, people lose their retreat
• Adjacent: G2.06 Global and local navigation · G2.10 Stability cost of personalized navigation · E5.18 Permission visibility of navigation
• Search terms: contextual navigation · predictable change · learned nav rules