Stable structure with variable content is easier to master than a structure that also moves

What it is

Home is always “search, continue, for you, browse,” and the titles inside can change every day — easier to learn than “some people see browse, some see discover, and sometimes search is inside a hamburger.” Stable structure, variable content means personalization should happen to objects in slots, not to whether a slot exists, what it is named, or where it sits. People master the rooms of the house, not this morning’s newspaper on the table.

Location memory is already unreliable at the content layer. Hand structure to personalization too, and even the retreat to rooms is taken away.

Why it happens

Skill acquisition depends on a repeatable skeleton. Each time the skeleton moves, old action sequences are void and people have to pathfind again. Content change can be read as “today’s paper is different”; skeleton change is read as “the house was rebuilt” — relearning costs an order of magnitude more. Personalize the skeleton as well, and every person, every day, is a small redesign.

Stable structure also leaves a layer for interpersonal directions: you can say “in Browse,” even if the items in Browse differ. Once structure moves, that layer of public language goes too. So stable structure is not a vote against personalization. It is personalization confined to content slots, so spatial memory and teaching can still rest on the skeleton.

Studying it

Two personalizations: swap only objects in slots, versus also swap which slots exist, their order, and their names. After a week of use, measure: task time (go to browse, go to search), drawing the home skeleton from memory, telling a stranger how to walk it. Independent variables: how much structure varies, whether users can pin structure. Dependent variables: skeleton recognition, task time, whether restatement can still use relative position.

Do not mash “freshness” and mastery into one score. A moving structure can feel lively and still leave people unable to draw the house.

Where it stops holding

A/B tests and rollouts will make structure differ across people; keep experiments in content slots, not in navigation. Adaptive complexity (an expert mode) is structure change the user chose, not silent personalization. On small screens structure is already folding; stable means the folding rule is stable, not every pixel. This entry chooses which layer may be personalised. It does not treat the need for a search fallback.

Applying it

• Write the navigation IA as a skeleton identical for everyone. Personalization fills content slots. New rails ship to everyone, not per person.
• Let users pin rail order. What is pinned is structure and must not be rewritten by the next sort.
• Check: have someone who has used the product for a month draw home from memory. Slot names and relative positions should be drawable; items inside slots may be blank. If even the slots cannot be drawn, structure has already been eaten by personalization.

Related

• Same group: L6.12.1 Personalization makes the same entry show different content, so location memory fails · L6.12.2 Content that changes over time makes previously seen items hard to find again · L6.12.3 Users cannot give each other directions; shared screenshots no longer match anyone else’s interface · L6.12.5 Search and history that are not personalized must exist as a fallback
• Nearby: L6.10 Turning Personalization Off and Resetting It · L2.02 Discoverability of What Can Be Said · L1.01 Mismatch Between Stochastic Output and Deterministic UI
• Search terms: stable structure, variable content · learnable personalization · information architecture