This dataset relates to a paper exploring the most data and energy intensive practices (watching, listening, social networking etc.) and time of use of this demand on Android devices (http://dx.doi.org/10.1145/3025453.3025730). This is a dataset arising from a mixed methods study involving detailed logs of application use on mobile devices, interview data and associated analysis. We include aggregate data for the Atlantic Archipelago dataset (398 devices); this data is used in the paper both textually and for the figures, and details about the data provided is listed below. To preserve participant anonymity, we are not at liberty to disclose data for our formative Eight Participants dataset. Instructions on how to access the Device Analyzer dataset are provided here: https://deviceanalyzer.cl.cam.ac.uk/.

Note that we do not analyse app use or demand, and therefore category use or demand, for any apps installed on less than 50 devices to maintain an anonymity set. To further maintain participant anonymity, categories which have less than 10 devices (i.e. less than 10 of the devices demanded data or used apps within a category) have been removed, and all values have been rounded to 3 decimal places. Values which are zero are represented as ‘-0.0’, and values less than 0.000 (i.e. effectively zero) are represented as ‘+0.0’.

Hourly data files:
All hourly data files listed below present the average time use (i.e. the number of times an app had foreground status when the device screen was both on and unlocked), or average data demand (in kibibytes), of devices who were ‘active’ devices within each category- i.e. they were seen to use an app within a category, or the app within a category demanded data. For this data, we average across days for each device and then average across devices. The number of devices active within each category is provided in brackets next to each category name.

1. AtlanticArchipelago_HourlyDataDemand.csv – this data represents the distribution of data demand hourly across the demanding app categories for Atlantic Archipelago. The categories of ‘Watching’, ‘Listening’, ‘Social Networking’ and ‘Communication’ were used to produce Figure 5b in the paper. The categories of ‘Background Processes and Services’, ‘Software and Application Updates’, and ‘Storage, Backups and Transfers’ were used to produce Figure 7 in the paper.

2. AtlanticArchipelago_HourlyTimeUse.csv – this data represents the distribution of time use hourly across the used app categories for Atlantic Archipelago. The categories of ‘Watching’, ‘Listening’, ‘Social Networking’ and ‘Communication’ were used to produce Figure 5a in the paper.

3. AtlanticArchipelago_HourlyCommunicatonUse.csv – this data represents the distribution of communication time use hourly for Atlantic Archipelago, used to produce Figure 6 in the paper. Hourly communication use consists of the number of: SMS sent and received, phone calls, and times communication apps were in the foreground.

Daily data file:
1. AtlanticArchipelago_DailyDataDemandTotals.csv – this data represents the total daily data demand from the 350 devices which demanded data from the Atlantic Archipelago dataset. For this data, we average across days for each device and then sum each device for the daily totals. This data was used textually in the paper. We provide the following values:
a. Overall demand (bytes): the total data demand each day for the Atlantic Archipelago devices.
b. Overall demand (bytes) from categories: the total data demand each day that we are able to categorise, and its percentage of overall demand each day.
c. Category demand (bytes): the total average data demand for each category, and each category’s percentage of overall demand.

Data files are embargoed until 6/5/2017.