Insight
Stable Frame, Dynamic Contents
If the interface is different every time, how does anyone learn it or trust it? The answer isn't less adaptation. It's splitting the interface into a frame that stays and contents that move.
The objection: different every time
An interface that rearranges itself on every visit. Spatial memory is lost, and in a system handling coverage or money, exhausting turns into untrustworthy.
Frame versus contents
The room stays — shell, primary navigation, the persistent landmarks. Only the work on the table adapts to the situation you came in for.
Two failure modes and the target
Static (frame and contents both fixed) and chaotic (everything moving) versus the target: a stable frame with dynamic contents.
The first thing people say when I explain situational interfaces is fair: if the screen is different every time, how does anyone learn it or trust it? Spatial memory is real. People rely on "the thing is always in the top-right." An interface that rearranges itself on every visit is exhausting, and in a system that handles coverage or money, exhausting turns into untrustworthy.
The answer isn't to be less adaptive. It's to split the interface into two layers. The frame stays put, always: the shell, the primary navigation, the landmarks. It's the room. The contents adapt: the work assembled for the situation at hand. Same room every time, different work laid out on the table depending on why someone came in.
Get the split wrong one way and the frame moves too, and nobody can find anything. Get it wrong the other way and the contents are fixed as well, and you're back to a static screen. The first version I tried let everything move, and even I couldn't find things in it.
The frame is the part people learn once. I try not to touch it.