TV ratings provider BARC aims to have 50,000 panel homes, make use of AI

While the agency is working on expanding the sample size and making the panel more representative of TV viewing India, it is also making use of AI and machine learning to help with analytics and data

With the Telecom Regulatory Authority of India (TRAI) recently floating a consultation paper on television viewership measurement in the country, the spotlight is on the Broadcast Audience Research Council of India (BARC). In operation a little under four years now, the ratings agency replaced TAM Media Research, a joint venture between Kantar Media and AC Nielsen, as the sole TV ratings provider in the country.

Questions have been raised over the adequacy of the panel size and whether the sample is representative of the actual population watching television across the country. The recent TRAI consultation paper hopes to address these and similar issues raised by stakeholders in the media and entertainment industry.

While BARC welcomes the consultation paper, and believes an exercise like this would “help us further strengthen and improve the measurement system.” The agency currently has 33,000 panel homes, which has grown from 10,000 at launch in 2015. The aim is to eventually have 50,000 panel homes.

“The current sample panel size is 33,000 households, which will soon go up to 44,000 households. In order to ensure the quality of the sample, and the fact that it is representative of the entire population, the Broadcast India Survey with a sample size of 300,000 households is conducted,” explains Derrick Gray, chief of measurement sciences, BARC India.

 

Apart from increasing the sample size and ensuring that the panel is representative of the TV viewing audience in the country at all times, the agency is investing in technology to increase the robustness of data. “We process viewership data from 33,000 metered homes every eight seconds, and then combine this data with the demographic diversity and the broadcast content. With increasing data volumes and hundreds of permutations and combinations, we quickly moved to a combination of Statistical Models (Machine Learning), SME knowledge base and local intelligence. With the adoption of AI and ML, the system can explain viewership spikes and highlight the outliers,” says Gray.

He gives the example, of Republic Day and Independence Day when there is a distinguished peak in viewership, which the system attributes to the respective event instead of flagging it as an anomaly. The agency is working on improving the machine learning capabilities in its software.

He further explains that the relationship between increasing sample size and robustness of data is not linear, especially in a country as diverse as India. The government had mandated that the ratings agency have 55,000 panel homes. “While many individuals will have different opinions as to what the exact sample size should be, it cannot be denied that BARC India’s goal of 50,000 is sufficiently large and BARC India is the largest television measurement panel deployed in the world,” Gray says.

India currently has the largest sample size at an individual level when compared to other major TV markets across the globe with a sample size of 135,000 individuals. China, which globally has the largest TV viewing population of 1.27 billion individuals, has a TV panel comprising of 22,500 individuals. Countries such as the UK and Japan have panels of 12,440 and 10,000 individuals respectively. The USA has a panel size of 108,900 individuals for a TV viewing population of 298 million individuals.

The challenge of TV viewership measurement in a country large and diverse as India are manifold. Gray explains that the diversity in itself is a big challenge. “Each state is heterogeneous, with multiple languages being spoken in each. We have effectively addressed these challenges by establishing a properly stratified and balanced sample, inclusive of all languages and regions.”

“Sample size and design are the two principal factors that lead to accuracy and precision of survey estimates, forming the core to what is called a robust measurement system. For BARC India, these two are areas of high importance and attention,” he adds.