120,626,225 targets. This equals 314 touch events per pixel on a state-of-the-art screen with a resolution of 480 x 800 pixels. The amount of data enabled us to show how touch positions are systematically skewed. On the basis of this data, we derived a function that shifts the users’ input as compensation. We evaluated the compensation function by publishing it as part of an update of the game and collecting data from an additional 12,201 players. The results show that the compensation function significantly reduces the error rate.

The approach enabled findings that can hardly be derived from traditional studies. When examining target selection on touchscreens, the results are affected by noise, even when using homogeneous samples and controlled lab environments. Facing a large number of confounding factors, we discovered, for example, that introducing a compensation function has only a small effect. But even an improvement of just a few percentage points becomes relevant if the task is as pervasive as selecting targets on touchscreens. However, a large sample is needed to verify this effect. The largest amount of target-selection trials ever collected enabled us to describe the users’ behavior very precisely. The amount of data enabled us to analyze the user behavior for all screen locations and a large variety of target sizes. In contrast to related studies, our dataset spans more than a hundred different smartphone models. Each additional factor that one considers, such as targets with different shapes and colors or different user groups,

 

• In our field experiments, we have shown that navigation instructions encoded in vibration patterns can significantly reduce distraction and allow people to focus on the environment.