BARB Explained: Turning device ratings into people ratings 24 August 2018 In September, BARB started to report multiple-screen viewing figures, showing the number of people watching programmes across four screens: TV sets, tablets, PCs and smartphones. To do this, we have to turn the device-based census data we collect into people-based viewing figures – but how? BARB has two sources of data. The first is device-based data for online TV viewing; we collect these whenever anyone in the UK watches television via a broadcaster video-on-demand (BVOD) service. The other source is our nationally-representative panel of UK households, which have software meters installed that track their viewing on TV sets, tablets and PCs. These two sources are combined using a process called Dovetail Fusion in order to produce multiple-screen viewing figures. As we measure viewing on all devices in the same panel, we have insight into how individuals view across different screens. There are two aspects of online TV viewing we need to understand from the panel: the extent to which people watch online TV programmes on their own or as part of a group (known as viewers-per-view) and what type of people watch online TV on different devices. The first step in producing multiple-screen viewing figures is to convert the device-based data into individuals. An all-individuals number is calculated by applying a viewers-per-view factor to the average programme streams. This factor is sourced from the BARB panel, which provides insight on the extent of joint viewing across different programme types and devices. There is one caveat to this; as we discussed in a previous BARB Explained, we don’t yet track how our panel members watch on smartphones. Due to this, we have initially assumed one viewer-per-view for each smartphone viewing session, in order to include smartphones in our multiple-screen viewing figures. This also means that smartphone viewing is only available for the all-individuals demographic group, until we have a reliable panel measure of viewing by different types of people on smartphones. The next step in the process is to apply demographic profiles. Firstly, we check which BARB panel members have viewed the programme in question on their devices. If the sample is sufficiently large, that demographic profile is applied to the census viewing. If not, then we take additional variables into account in order to generate a larger sample, such as viewing to the same programme in the previous week, or of viewing to other programmes in the same genre. We can then report a comprehensive output of audiences to every programme viewed on four screens each day. This will be followed by more detailed cross-platform reporting, which will enable the calculation of incremental reach and time spent viewing. In the meantime, check out this previous BARB Explained to find out what we already know about online TV viewing.