Viewing data

Update: Project Dovetail

26 April 2016

Project Dovetail has become synonymous with BARB’s development strategy. The objective is to deliver robust viewing data across platforms and devices.
The strategic context is set by the increasing fragmentation that is brought on by the growth of timeshift viewing, new on-demand services, content distribution through the internet and the advent of dynamically inserted advertising.
Here are six things to know about Project Dovetail.

1 Hybrid measurement is the future. Collecting data from devices that are used to watch television is a cost-effective way of building large samples. Yet viewing information from devices doesn’t tell you who is watching: this limitation is balanced by the strengths of BARB’s panel of 5,100 homes. The viewing information collected from our panel delivers programme reach, demographic viewing profiles and measurement of viewers per screen. Project Dovetail will harness the complementary strengths of device data and panel data.

2 Generating census data. BARB is now collecting data from software code that has been included in TV player apps such as All 4, BBC iPlayer, ITV Hub, Sky Go and UKTV Play. Each software implementation is independently validated and audited before BARB publishes the viewing levels; these are generated from across the whole population, not just from a nationally representative panel of homes. The data are published each week in the TV Player Report and feature in the Trend analysis section.

3 New metrics are needed for online TV. It’s important to report the average duration audience for programmes and commercials to provide comparability with established TV audience metrics. BARB
already reports average programme streams in the TV Player Report and plans to start collecting average ad streams in the coming months. Each of these new metrics, which have been ratified by JICWEBS, is a measure of device usage, rather than than a measure of how many people are viewing.

4 Device data need to be converted into people data. The hooks for creating a data fusion will come from BARB’s panel of homes: we have installed software meters on the personal computers and tablets of over 2,000 homes on our panel. This is generating observations of how people are watching online TV, insight that will be critical to the data fusion process.

5 Data fusion is not straightforward. While not a new concept, our aim is to provide a fresh data fusion every day. BARB also has to ensure that our chosen method stands up to the scrutiny of the television and advertising industry. To this end, we have commissioned two research agencies to deliver prototypes of their proposed solutions. We expect to complete our review of these approaches during the second half of 2016.

6 Other data sources may be available. Set-top box data offer similar benefits to the information BARB is collecting from TV players. We don’t envisage sourcing census counts from set-top boxes, although working with large samples would achieve a number of objectives; an increase in effective sample sizes, reduction in data variability and providing certainty on the identification of new forms of dynamically inserted advertising. BARB has successfully conducted pilot projects with data from Sky homes and expects to complete further development work during 2016.