Across the world, efforts are on to make audience measurement systems more reliable and tamper-proof so that advertisers can get the right value and content providers can tailor content as per market demand.
Return Path Data (RPD), the process India's audience measurement body BARC will soon implement to count viewership, involves TV viewing data of homes with addressable set-top boxes (DTH and digital cable) by enabling “return path” flow of data.
RPD is being used by distribution players to study consumer behaviour across the globe, including UK, US, Canada, South East Asia. But nowhere it is a part of the viewership measurement currency and BARC India is aiming at that.
To understand the benefits of RPD and how it is going to affect India's broadcasting industry, BestMediaInfo spoke to Ricardo Gomez-Insausti, Vice-President Research, Numeris. Ricardo, a global expert in audience measurement system, is currently, pioneering the RPD implementation in Canada. He will address a session on RPD on Wednesday at MELT 2018 in Mumbai.
Other than getting a tamper-proof data, he said RPD will make the TV networks catering to niche markets benefit from the data granularity that RPD can offer. "It is a step closer to dealing with targeted advertising within a highly segmented market," Ricardo said.
Speaking on why the existing measurement body was best suited to implement RPD, he said, "The integration process of RPD with panel-based data requires rigorous empirical work that needs to be done progressively. This is why it is felt that the existing measurement company is best equipped to undertake this."
Allaying privacy concerns related to RPD, Ricardo said, "RPD is a sample-based measurement, but a much larger sample than a meter-based one. And if the data is sourced from a sample, and not the universe, privacy concerns are largely allayed."
He said regulators could also ensure that the detailed information on STB subscribers is not shared. "The objective would be to reach a level of anonymisation that is good enough to do the data fusion without compromising the usability of the outcome," he said.
Ricardo added that just RPD wasn't enough for a substantial market representation. "RPD data used for viewership projections may need to be statistically calibrated by panel data to guarantee reasonable market representation," he said.
Based on your wide experience in the audience measurement, do you feel the requirements of audience measurement has changed and become more important in the last 10 years?
Definitively. The fragmented consumption of media requires innovative approaches to measure audiences anytime and anywhere. Embracing hybrid approaches to measurement is the way to go. The integration of panel data with subscription-based and machine-generated data is one of the appealing strategies. This is the time for innovation and thinking outside the box in audience measurement! I am looking forward to addressing this question in further detail in my keynote at MELT 2018 on Wednesday.
Everyone is obsessed with data these days? How much of the data that is available is actually being deployed intelligently for business insights? Is this improving over time?
Yes, it does. However, we still face the challenge of understanding how data and information are created before using them or deploying them to make business decisions. Data integration methods are also able to create new sources of information that need be clearly understood before coming too quickly to conclusions. The act of balancing quality, speed and cost in audience measurement is key for decision makers to understand before making decisions and accepting trade-offs.
Census is the holy grail for any measurement. Is RPD a halfway house towards that? How does RPD improve panel-based data?
Return-Path-Data provides an incredible amount of viewing information that can complement panel-based data. Yet, this phenomenal information is originated by a TV set-top-box, which ultimately produces device-based data that need to be understood and treated carefully. RPD can provide a level of data stability and granularity that is critical for doing business with a highly fragmented viewership.
India, with all its diversity and fragmentation, seems to be an ideal market for RPD-based TV measurement. Your take on that.
I completely agree, given the magnitude and diversity of the Indian population. But it is not just the Indian case. The cultural and social diversity of large metropolitan areas is a global phenomenon that deeply affects media consumption. For example, in Toronto, with just over 5 million inhabitants, we have more than 140 languages spoken. The TV networks catering to niche markets benefit from the data granularity that RPD can offer. RPD is also a step closer to dealing with targeted advertising within a highly segmented market.
Globally, how many countries use RPD for measurement?
There are several markets that have explored / integrated RPD for measurement purposes. For example, the UK with Sky, USA with Dish, Direct TV, Comcast, France with Canalsat and South East Asian countries with the likes of NowTV, Astro, StarHub and SingTel have tested and/or integrated to various degrees RPD into their measurement systems. The integration process of RPD with panel-based data requires rigorous empirical work that needs to be done progressively. Which is why it is felt that existing measurement company is best equipped to undertake this. In Canada, we are in the intermediate stages – and hope to integrate RPD for measurement.
What are the downsides of RPD-based measurement? What would you say are the red flags that need to be kept in mind here?
In the case of Canada, RPD might only be representative of about 60% of the market that has set-top-boxes, which are not always capable of returning data. Therefore, the geographical and social representation of the RPD homes may be limited. Since local suppliers and STB types vary considerably by country, region and even metropolitan areas, RPD data used for viewership projections may need to be statistically calibrated by panel data to guarantee reasonable market representation.
Following from that, do you feel audience measurement bodies like Numeris are best placed to also undertake RPD-based measurement?
Transparency and trust are extremely important in order to support the use of currency. Like BARC India, Numeris is a non-for-profit tripartite organisation governed by radio and TV broadcasters, agencies and advertisers. They review all results and have direct input through research and executive committees, and finally the board. Additionally, our TV and radio PPM services have been audited by EY following the US Media ratings Council’s standards. Due to this robust and transparent governance structure, a third-party approach by the currency provider is best suited to roll out RPD-based measurement.
How enabling has technology been for audience measurement?
Numeris top two priorities are enhancing the measurement systems that rely on the integration of digital data from online video consumption, and RPD from set-top-boxes. Both types of technologies facilitate the development of a hybrid approach to measurement. The enlargement of the Portable-People Meter panel needed to support the measurement of the highly fragmented viewership is not feasible. Our goal is to use the high quality panel data to inform and calibrate the data coming from machine-based sources. In this context, the technology behind the data integration methodology is key for the expansion of the measurement system.
There is a view that when industry moves from the diary system to technology-based measurement, viewership/ listenership drops. Why is it so?
I would agree as a generalisation. However, we cannot forget that every data collection tool has its own intrinsic limitations. At Numeris, we have faced general declines in audience estimates when we moved from diaries to any of the electronic technologies we have used over time (i.e., PMT, LPM, PPM). Our diaries collected respondents’ recall data by quarter hours, while the meters collected respondents’ active or passive exposure to media by time units smaller than a minute. In comparative analyses of both types of data from the same respondents, we have observed numerous discrepancies, particularly in times of well-established daily habits (e.g., early morning radio, TV news). At the end of the day, the most important aspect is the supply of evidence and rational explanations for the differences introduced by the methodological change.
Research always draws sharp and often opposite reactions: if numbers look good, everyone likes it. But when they don’t, people tend to criticise the methodology, etc. How do you react to those sentiments?
I do know the feeling. I believe the best approach is to be transparent while sharing test results and simulations, and providing as much information as possible to appraisers and users. At the same time, there is never total guarantee of the outcome since we deal with sampling. It is important, however, to develop a detailed educational plan to tackle the business impact of any proposed methodological change. My role is to ensure that our estimates are produced by a sample that properly represents the market to the best of our capabilities.
You are also involved in radio audience measurement. The Indian radio industry is small in terms of ad spend, but very vibrant in terms of number of players, audience size/spread, etc. What would you suggest would be a workable measurement system for the Indian radio industry?
The challenge is to have an affordable research approach that can support the radio business at the market level. Our measurement is not the same in all radio markets. The radio business is very local in Canada. The top markets are measured electronically while the smaller markets are not. The fact that we use a TV-radio PPM panel with standardised metrics (Average Minute Audience) diminished the costs for both industries in the top markets. Diaries are still used for the remaining radio markets.
Data sharing and privacy standards have come into sharp focus recently. How does that impact RPD-based measurement, if at all.
From what I have read, and from my limited interactions so far, it seems some sections of the industry and media in India believe RPD is the same as census measurement. It is important to reiterate that RPD is not census. RPD is also a sample-based measurement, but a much larger sample than a meter-based one. Each operator contributes a certain sample, which is a subset of his subscriber base. And if the data is sourced from a sample, and not the universe (i.e., census), privacy concerns are largely allayed. Of course, based on the level of data integration, some privacy concerns may remain. That’s where the regulatory environment comes into play. For instance, in Canada, under the current environment, detailed information on STB subscribers cannot be shared. The objective would be to reach a level of anonymisation that is good enough to do the data fusion without compromising the usability of the outcome.