UK TV measurement body Barb this morning announced the launch of a new prototype phase for Barb Panel Plus, its initiative to integrate big data sets with its classic panel measurement. Two prototypes — one from Kantar and one from RSMB and Sopra Steria — have been commissioned, and Barb will test both before making a decision on how to proceed by the end of the year.
Two big datasets are being tested within these prototypes: return-path data from 900,000 Sky set-top boxes, sourced by measurement business TVbeat, and return-path data from HbbTV enabled connected TVs in 26 million homes, sourced by TVA.
Both prototypes will integrate these two datasets with the existing data sources Barb works with, namely its panel-based data and data it receives about viewing on broadcasters’ streaming services. The aim is to get the best out of all these data sources, combining the scale of the return-path data with the robustness and demographic insights provided by the Barb panel.
“Big data provide a precise measurement of the volume of viewing; while panel data provide valuable insight into who and how many people are in front of the screen,” said Caroline Baxter, chief operating officer at Barb. “By integrating further new sources of big data, Barb Panel Plus will enhance our audience measurement by improving data consistency, reducing zero-rated spots and facilitating insight into business outcomes for our clients.”
It’s a match!
Barb Panel Plus isn’t the company’s first foray into big data. As mentioned, the measurement body already uses census-level BVOD viewership data collected from smartphones, tablets, and PC, through its Dovetail Fusion project.
Dovetail Fusion was launched to capture viewing happening outside of TV sets. Barb Panel Plus meanwhile is designed to improve the granularity of viewing happening on TV sets — which has become more and more important as viewership has fragmented across devices and services.
Combining big data and panel-based data however is no mean feat. There are plenty of big data sources covering TV viewing, thanks to the growth in internet-delivered TV and smart TV devices. But these datasets differ in their levels of reliability, reach, granularity, and whether they include any further information on who exactly is watching. Barb’s two choices suggest it has decided against using automatic content recognition (ACR) data, despite its growing role in the measurement landscape.
Once big datasets are chosen, they then have to be matched to panel data — this is where Kantar and RSMB/Sopra Steria come in. Return-path datasets are large but also fairly raw. They don’t necessarily tell you much about who was watching, and they’re unlikely to be representative of the entire population. So the prototypes essentially have to use the information they get from both datasets to make judgements about viewing across the whole population — a process which involves lots of complicated maths and plenty of judgement calls.
The hope though is that this work results in an overall more accurate way of measuring TV, in turn helping drive more ad spend back towards broadcasters.
Follow VideoWeek on Twitter and LinkedIn.
