A dialog's close button can be engineered to be hard to hit
Aliases: tiny close button · misdirected tap target · close trap
What it is
The physics of the close button can be used in reverse: place it in the thumb's blind zone, shrink it below reliable tap size, offset the hit area from the visual area — "closing a dialog" gets designed as a retention tool. Mobile thumb-reach maps give this design precise engineering means, and precisely because of that, precise detection means too.
Why it happens
Tap cost follows Fitts's law: acquisition time grows with distance and shrinks with target size — used in reverse: small target plus far position equals high abandonment. The technique list: area (the close X shrunk to pixel-scale), position (screen's top corners, outside one-thumb reach), hit mismatch (visual bounds larger than hit bounds — the tap looks landed but misses), latency (response delayed half a second, inducing repeat taps or giving up), and fake close (tapping X raises "are you sure you want to leave?"). The conflict with accessibility standards is explicit: minimum touch-target size is a platform-level norm, and a deliberately shrunken close button violates it directly — which makes this the most automatable dark pattern class to detect.
Where it stops holding
Small close buttons have other origins: minimalist design taste, or hit-area configuration accidents — visually identical to deliberate masking, so judgement reads consistency: if every touch target in the app meets spec except the close button, intent is clear. Pre-roll ad video's "no close button for the first seconds" is industry convention, but convention does not change the mechanism — judge by mechanism, not by convention. And a positioning tradition confuses things: top-right close is desktop convention, so the position alone is not evidence of manipulation.
Applying it
- Spec it: every dialog's close button carries a hit area no smaller than the platform norm (44pt/48dp), positioned inside thumb reach, or the whole dialog closes on outside tap.
- No second-chance retention on the close path: "are you sure" dialogs belong to the marketing module's separate review and may not hang off the close event.
- Verification: automated touch-target scanning — traverse all dialog components, list every undersized hit area, and gate releases on the fix rate.