Stable core nav and personalized modules must be partitioned and must not invade each other

What it is

Core nav (classes, roads to main tasks) owns a fixed region; personalization (recommended, recent, for you) owns another, and the boundary does not stretch with recommended content. Partition with no invasion: extra recommendations cannot push Settings out of the core, and core classes must not be ranked by an algorithm inside the recommended module. The two blocks may sit next to each other; they must not share one stretchable slot strip. When invasion happens, both the stability promise and the personalization promise go bankrupt—the stable region starts to jitter, the personal region starts to impersonate structure.

Partition is a layout contract, weaker than “both sides have links.” The contract has to point at which pixels are core and which are personal.

Why it happens

People pick a memory strategy by region: core by location, personal by scanning content. A fuzzy boundary picks the wrong strategy—pointing at recommendations, rereading core items every time. If the recommended module expands into the core (one more recommendation eats a class slot), core width becomes a function of the recommendation quota and spatial memory has no fixed frame. If a core class is written into the recommender, the same object appears in both regions; people have to choose which to click, and location memory is copied apart.

Visual partition (a rule, a background, a “For you” heading) is the cue for strategy choice. Without the cue the two blocks are one array in perception, and invasion is experienced even when it has not happened in data.

Studying it

Manipulate boundary clarity and whether invasion is allowed, and test whether the two memory strategies stay in their lanes.

• Paradigms: five fixed core items + three recommendation slots; one condition lets recommendations insert into the core, one hard-partitions. Mix “go to Settings” with “open today’s recommendation.”
• Independent variables: presence of a For you heading and separator, whether recommendations can change core width, whether the same object appears twice.
• Dependent variables: whether core pointing still hits fixed coordinates, whether recommendation tasks mis-hit core slots, whether people can later sketch two regions.
• Methodological note: measuring only recommendation clicks rewards invasion (recommendations further left are easier to hit). Core tasks must be measured too. At responsive breakpoints the two blocks may wrap and the boundary vanish; retest at those breakpoints.

Where it stops holding

A small screen may fit only one nav strip; partition then becomes “core owns the strip, personalization goes to overflow or a separate page,” not two truncated blocks in one strip. User-pinned items are legal members of the core, not recommendation invasion—the pin is human-initiated and the place is human-set. A campaign occupying a core slot is editorial invasion with the same failure mode as recommendation; governance should use the same “no invasion” rule.

Applying it

• Draw two non-stretching regions: core nav has a fixed slot count; the recommended module has its own fixed slots or a separate card, with overflow handled inside the module.
• Core items are forbidden on the recommender’s insert list, so they do not appear twice.
• Verify by filling recommendations from zero to capacity: core items’ screen coordinates should not move. If they move, they were invaded. Then ask someone to circle “won’t change” and “will change” on a screenshot; inability or a wrong circle means the partition does not hold in perception.

Related

• Within the group: G2.10.1 Reordering nav items by frequency unfixes their positions and hurts predictability · G2.10.2 Personalized recommended items appearing and disappearing interrupt spatial memory · G2.10.4 Personalized ranking belongs on secondary entrances, not primary-task nav · G2.10.5 People need an explicit way to turn personalization off and restore a fixed order
• Adjacent: G2.06 Global and local navigation · G2.07 Navigation consistency · F2.06 Grouping by proximity
• Search terms: nav partitioning · stable versus personalized · spatial memory