Seeing a component up close doesn't help unless the operator still knows where it sits in the process
Aliases: overview-detail coordination · industrial human factors
What it is
Overview-detail coordination keeps an operator's sense of where a selected component sits in the larger process while they inspect it closely. Putting an overview panel and a detail panel on the same screen is not enough on its own — the two views must map onto each other, so selecting an object in the detail view makes its position and reach visible in the overview, and vice versa. Without that correspondence the two panels are just two unrelated sources of information sitting side by side.
Why it happens
Diagnosis needs exact values, exact wiring, exact thresholds — that is local precision. Deciding whether a fault will propagate, and whether it needs attention now, needs the surrounding structure — that is global context, roughly what Endsley's three-level model calls comprehension and projection sitting on top of perception. Neither judgment substitutes for the other, so a display that only supplies one leaves the other judgment unsupported.
The cost of moving between the two is not the act of switching screens; it is the working-memory step of re-registering "where was that object" once you are back in the overview. How expensive that step is depends on whether the two views share a common frame of reference. If overview and detail are drawn on the same spatial layout with the same object identifiers — say, the same process diagram at two zoom levels — switching is a lookup, and linked highlighting closes the gap cheaply. If the two views use incompatible geometries — a plant floor-plan overview against a schematic loop diagram with a completely different shape — the operator has to mentally transform coordinates every time, and synchronized highlighting does not remove that cost, because the problem is not "can't find it" but "can't translate it." Whether linked interaction can substitute for coordinated design turns on this single condition: do the two representations project onto the same frame or not.
This also explains the trade-offs among the three common techniques: a split-screen layout needs no coordinate transform but consumes screen area; focus-plus-context (fisheye-style distortion) keeps everything in one frame but the distortion itself interferes with precise reading; separate, paged views have the highest switching cost but scale to systems of any size. The choice is set by whether a shared frame exists, plus available screen area and system scale — not by preference.
Studying it
Compare split-screen, focus-plus-context, and separate-page designs on fault-location time, cross-level backtracking, accuracy on global-state questions, and wrong-object actions. The task must require both local diagnosis and a judgment about system-wide effect at once; a task that only asks participants to "find the reading that changed" measures local search speed, not overview-detail coordination, and a design that wins on that task alone tells you nothing about safety during a real fault.
Where it stops holding
Small screens, handheld field terminals, and very dense process diagrams cannot keep both layers visible continuously; forcing them side by side just shrinks the readable area for the detail that matters. On such devices a single primary view backed by on-demand lookup usually beats a forced two-pane layout.
Coordinated interaction cannot fix a representational mismatch: if the detail view shows something the overview never modelled — internal firmware state or register values, for instance — there is no position on the overview to highlight, and no amount of linking closes that gap. The fix is to add a placeholder for that object in the overview, not a better interaction technique.
An overview is also not ground truth. Aggregation can average away a small but severe local anomaly, so relying on the overview alone for risk judgment carries this same limitation.
Applying it
Keep the overview continuously showing the operator's current location, the affected scope, and key global states — not just on first load. Synchronize selection highlighting between views, and link time cursors across them so both sides are never showing different moments.
Before choosing a technique, check whether the two views share a frame: if they do, prefer split display with linked highlighting, since it is cheap; if the geometry differs, a plain locator cue (a pin on a mini-map) beats an automated visual distortion, and is less likely to be misread.
How to check: run cross-scale fault drills and record both the time to locate the local fault and the accuracy of the operator's answer to "what does this affect" immediately after. Both numbers need to drop relative to baseline before you can call the coordination effective — if locating gets faster but the global-impact answers do not improve, the change only sped up navigation and never solved coordination.