Hit shapes should follow where fingers actually land

What it is

Fingers do not land in a uniform halo around an icon’s centre. A one-handed thumb tends below and inward; an index tapping a small close-box tends below and right; a corner button has its top edge occluded by the palm. A hit region that only scales the visual outline up enlarges pixels people rarely hit above the glyph, and still misses the dense landings below. Shape should follow the landing distribution—usually a one-sided ellipse or rounded trapezoid—not a Minkowski dilation of the outline.

Why it happens

Aiming is shifted by joint range, occlusion, and the systematic bias of “I meant the top of the icon but I am looking at my fingernail.” The distribution therefore has both width and offset, with a major axis often along the thumb’s sweep. Uniform expansion adds isotropic area and does nothing to the offset: the centre still misses the distribution mean, one side keeps missing, the other starts stealing from a neighbor. Bottom navigation, search-field clear buttons, and narrow keyboard keys show the largest offsets. Drawing a hot zone from visual symmetry assumes isotropic noise, which touch is not.

Studying it

Under real grips, collect a 2-D histogram of contacts relative to each target’s visual centre, fit mean offset and a covariance ellipse, then compare three hot zones: uniform expansion, an ellipse-placed region, and a region that extends only downward. Independent variables are grip (left / right / two-handed), screen partition, and finger (thumb / index). Dependent measures are hit rate, neighbor false hits, and extra hot-zone area that is never tapped. A lab phone lying on a desk poked with an index yields an overly symmetric distribution and erases the evidence for offset.

Where it stops holding

Mouse and stylus landings sit closer to visual centre; shaping to a finger distribution makes the hot zone feel “crooked” to an indirect pointer. A public kiosk mixed across users and grips is hurt by a single offset; uniform or per-zone switching is more stable. Legal checkboxes and close buttons whose hot zones reach far into empty glass become dark hits. The distribution also moves with cases, films, and hand size; a heatmap from young right-handers on a flagship will invert on an elderly left-handed budget phone.

Applying it

• For frequent small targets, record a landing heatmap in the product’s actual postures and place the hot zone on the 95% ellipse, allowing obvious downward/inward extension rather than the same pixel padding on all four sides.
• Give bottom bars and keyboard keys different shapes by zone: edge thumb keys extend in and down, centre keys stay more symmetric.
• Split verification into top-edge misses versus neighbor theft; overall hit rate averages away “enough below, stealing above.”

Related

• Same group: C2.20.1 Hit regions need not match the visual shape of a control · C2.20.2 Irregular visuals are often hit-tested as bounding rectangles · C2.20.3 Bounding-box simplification can make neighboring irregular targets overlap
• Nearby: C2.06 Touch offset and aiming bias · C2.11 Thumb reach zones
• Search terms: landing distribution · touch offset · asymmetric hit target