Personalization makes the same entry show different content, so location memory fails

What it is

People remember “sport is the third rail on home,” “settings are top right.” Once personalization rewrites what sits inside the same entry, the third rail is no longer guaranteed to be sport; top right may still be settings — location memory at the content layer fails first. Personalization versus location memory means the same navigation coordinate points at different objects for different users, and even for one user at different moments, so spatial memory loses predictive force.

Not finding yesterday’s item in time is the next layer. What collapses first here is “what lives where inside this entry.”

Why it happens

Location memory depends on a stable map: coordinate → object. Personalization turns the object into “the set now best for you,” and the map changes every refresh. People still reach by the old coordinate and grab another set. The miss is felt as “I remembered wrong” or “they redesigned,” rarely attributed to ranking. Users then stop investing in location memory and scan from scratch each time — personalization spent a spatial channel that could have been free.

If structure (number of rails, where search sits) still holds, memory can retreat to the structure layer. If structure also varies by person (some have Discover, some do not), there is no retreat. Stable structure with variable content is a more learnable state, treated later on its own. Here the claim is only: content-layer personalization is already enough to break location memory.

Studying it

Let people learn coordinates for a week on a non-personalised layout, then switch to a personalised home. Measure: pointing with eyes closed to “where is sport,” pointing time, attribution after a miss (self / redesign / recommendation). Independent variables: whether personalization shuffles rail identity, whether it shuffles order inside a rail. Dependent variables: coordinate recognition, whether gaze paths turn from “reach” to “full-page scan.”

Control redesign. If a control only changes skin and not ranking, and memory holds, the failure can be booked to personalization, not to visual novelty.

Where it stops holding

Tool entries (search box, account) are usually not personalised, so location memory still works; failure concentrates on content rails. Expert users switch to search, which looks like adaptation and is actually abandonment of the spatial channel. On small screens there were never stable coordinates (an infinite stream); personalization is not the main cause. This entry is only content coordinates rewritten per person. It does not treat screenshots that cannot be shared, and it does not treat the need for a non-personalised search fallback.

Applying it

• Give content rails an identity that does not vary by person (name, icon). Personalize items inside the rail, not “what this rail is.”
• Do not add or remove whole rails per person. If they must change, go through explicit editing or user choice, not silence.
• Check: ask long-time users “without looking, where is sport.” If after personalization ships the answer moves from “third rail” to “I have to scroll to know,” location memory has already failed. Lock rail identity and test again.

Related

• Same group: 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 · L6.12.4 Stable structure with variable content is easier to master than a structure that also moves · L6.12.5 Search and history that are not personalized must exist as a fallback
• Nearby: L6.05 Turning Personalization Off · L6.10 Turning Personalization Off and Resetting It · L1.02 Expressing Capability Boundaries
• Search terms: personalization versus location memory · spatial memory · stable entry points