Compare gesture and non-gesture paths on discovery, entry, recovery, and exit—not only the successful stroke

What it is

Comparing mid-air gestures with buttons, touch, or a remote by timing only a single successful stroke yields the wrong answer. A complete use has four other legs: discovery (how does anyone know the movement exists), entry (how does the system start listening), recovery (what happens after a miss or a false fire), and exit (how does listening stop). All four must be timed, error-counted, and learning-counted before the path sits on the same cost sheet as a key press. Demos usually show only the middle, successful recognition.

Why it happens

A button makes discovery a visible control, entry a contact, recovery another press, and exit a release or a focus change. Mid-air lifts those four jobs off physical controls and onto memory, pose, and timeouts. Discovery is expensive because the action is invisible: nobody sees a button in the air. Entry is expensive because a signal must be separable from ordinary motion, or passers-by get harvested. Recovery is expensive because there is no obvious undo key; a false fire may already have skipped a track, turned a page, or killed a light. Exit is expensive because the system does not know the person has turned away and will keep treating later motion as input. Drop those four legs and gestures look cheaper than they are.

Studying it

Split a successful command into discovery, entry, execution, recovery, and exit timings instead of reporting only the execution slice. Independent variables include presence of instruction, complexity of the entry signal, recognition error rate, and whether exit is visible. Dependent measures include time to first success, help look-ups, time to recover from an error, and whether the system is still listening after the task. A laboratory that demonstrates first then times the trial zeros out discovery—the gap between the demo and the field. Field logs should also count extra recovery caused by bystanders entering the sensing range.

Where it stops holding

Experts amortize discovery and entry; their execution slice may beat walking over to a distant button. Using expert data to argue for a public device hides the unamortized cost. In games or performance, discovery can be the fun, so teaching time is not pure waste. If the gesture path has no equivalent non-gesture path, the comparison table is missing a column—that is not a gesture win. A very short timeout lowers exit cost and misreads a pause mid-execution as leaving.

Applying it

• In reviews, keep five columns—discovery, entry, execution, recovery, exit—and fill time, errors, and learning for both the gesture path and the button/touch path. Do not fill only execution.
• Run first-use with people who have not seen the lesson; separately record how long they take to find the action, confirm they are being listened to, undo a false fire, and leave with the system stopping.
• If gestures are clearly more expensive on one column, add visible cues and undo, or take that function off the gesture path. Do not decide from one successful recognition in a demo reel.

Related

• Same group: C4.01.1 Without contact, input is segmented from continuous motion · C4.01.2 Fit: contact is undesirable, distant control, or spatial objects · C4.01.4 Gestures are a poor fit for precise numbers and long text
• Adjacent: C4.15 Conditions for entering engagement · C4.28 Necessity of non-gesture paths
• Search: gesture discoverability · engagement cost · error recovery