5.2.1. Battery

Optimizing battery consumption is key to adoption of smartphone-based telemonitoring applications [39,45]. In this study, the subjects did not report any usability issues regarding battery consumption. However, as described earlier, battery levels were collected during every run (once a minute). From these data, we calculate the amount of battery that is consumed during each run. Note that the battery consumption captures the total battery consumption of the smartphone, not just of our app. Since the different runs are different in duration, we normalize this consumption by time, in order to get a measure of "battery percentage consumed per hour". These results are depicted in Figure 8a.

We note that different subjects have different battery consumption profiles, which may be caused by one or a combination of the following factors. First, different phones have different battery consumption profiles; and the phones used in this study were not provided by us and therefore are heterogeneous. Second, battery consumption depends on factors that are external to RunningCoach, such as listening to music and whether the music is streamed or played locally. Third, battery consumption profiles depend on the carrier and the strength of cellular coverage [46]. However, the data provide evidence that RunningCoach alone is not very burdensome on battery and can consume as low as 5% battery/h. This is even true for subject b01k1o, who manifests high battery consumption patterns in general. In one run, RunningCoach consumed less than 5% battery/h, which can be explained by the factors listed above.

**Figure 8.** (**a**) The amount of battery consumption (in %) per hour during the different runs by the different subjects (*N* = 22); and (**b**) the perceived accuracy of the collected data by the runners (*N* = 22).
