Discussions
Clarification for Variables in the App Usage Data
yesterday by Alex Wang
Hello,
We are currently working with the U.S. Mobile App Engagement dataset and would appreciate clarification on several key variable definitions related to events and usage time. During our data validationm, we observe unexpectedly large daily usage time totals after aggregating events at the panelist-day level. For example, summing durations sometimes produces totals that exceed 24 hours per day (even for VISIBLEDURATION). So we would like to kindly request your help with the precise definitions of events and activity windows to ensure that we compute usage metrics correctly.
- Each row in the dataset appears to correspond to an “event“ (FACTID) while including 3 other measures of engagement: foreground, visible, screen on/ff.
- How are these 4 concepts defined? For example, Does “foreground” correspond strictly to the Android foreground activity (top of the task stack), or can it also include foreground services (e.g., background audio playback)?
- At what levels are they defined? For example, is foreground defined at app level while screen on / off defined at phone usage level?
- How do they relate to / mapped to each other (event, foreground, visible, screen on/off)?
- How should HASCONCURRENTEVENT be interpreted?
Does this indicate true simultaneous app visibility (e.g., split-screen or picture-in-picture), or can it also reflect overlapping telemetry records during app transitions?
- What is the recommended way to compute app usage time?
For example, when calculating the app usage at the user-day or device-day level, should usage be based on:
- VISIBLEDURATION
- FOREGROUNDDURATION
- Or do we need to include other adjustments on top of the duration variables?
- Any clarification or recommended aggregation approach would be greatly appreciated.
- In the data, we also observe a substantial number users have zero screen time days (and there can be multiple such days for a given user). Why would that happen? Does it imply a data collection error from these users? Or do people use spare phones do participate in data collection and then do not turn these phones on sometimes?
Thank you very much for your help!