Your reader's eye groups things before their mind reads them. Five grouping principles do nearly all of the work in a diagram, and once you know them, "this looks cluttered" turns into a specific diagnosis you can fix.
What are the five Gestalt grouping principles?
Proximity is the strongest. Objects close together are read as one group, and the eye judges closeness relatively. It weighs the gap within a group against the gap between groups. Call that ratio the proximity ratio. Below about 2 the grouping is ambiguous, and the reader has to work it out from the labels, which is the very work the diagram was supposed to save. At 3 or above the grouping is instant and involuntary.
Similarity comes next. Objects that share a shape, a fill or an outline weight are read as the same kind of thing, so mixed shapes are expensive. A diagram with rectangles, rounded rectangles, ovals and hexagons has told the reader there are four kinds of thing. If there are really only two kinds, two of those shape choices are lies the reader has to unlearn. Pick one shape per category and never more than three shapes on a page, and make the shape difference mean something you could state in a sentence.
Advertisement
Enclosure is the most under-used and the highest-yield. A light box drawn around three nodes says "these three belong together" more clearly and more cheaply than three arrows do. You can see the arithmetic in a worked repair of a nine-node lending-marketplace concept diagram. The original carried 9 boxes, 14 arrows, 9 icons, 3 stray labels, a legend, 9 drop shadows and a title: 46 objects. The repair kept the 9 boxes and the title and cut the arrows to 5. It deleted every icon, label, legend and shadow, and it added 3 enclosures, for 18 objects in all, a reduction of 60.8696 percent. Three enclosures replaced nine arrows, and the proximity ratio moved from 16.0/14.0 = 1.1429 to 32.0/8.0 = 4.0000.
Continuity and closure are quieter, and mostly they act as constraints. Continuity says the eye follows a line and prefers a smooth path. An arrow that bends twice to dodge an obstacle gets followed less reliably than one that runs straight, and a diagram whose arrows all run left to right is read faster than one whose arrows run in four directions. Closure says the eye completes a shape that's nearly closed. So a dotted enclosure works as well as a solid one and costs less ink, and a box missing one side still reads as a box.
Proximity and enclosure cost nothing to draw. Arrows cost the reader an inspection each, and the budget is smaller than authors think.
How do Gestalt principles help you cut clutter?
The nine-node repair also shows what restraint buys. Nine nodes admit 36 undirected pairs. The original drawing asserted 14 of them, 14 relationships for the reader to hold, and the repaired version asserts 5 plus 3 group memberships. Nothing true was lost along the way. The nine arrows that went were real relationships that didn't matter to the claim the diagram supported, and each one had been costing the reader an inspection. Against the seven-rule slide check, the original scored 0.43 on the machine and 0.50 on the rubric. Under the 0.6 machine plus 0.4 rubric blend that's a composite of 0.4580, which misses the 0.80 pass by 0.3420. The repair scored 0.94 and 0.90, a composite of 0.9240.
So the habit to build is subtraction. Before you add an arrow, ask whether proximity or enclosure could carry the same relationship, since both are free and an arrow isn't. Before you add a shape, ask whether it names a category the reader needs. And when a diagram feels crowded, count its objects. The count is usually higher than the author believes, and 46 for nine ideas ends any argument about whether it's crowded.