Innovations and media audience measurement

Audience le mag

Big Data, artificial intelligence... the current momentum in innovation offers so many opportunities to enrich media audience measurement. Médiamétrie conducts a great deal of R&D in these fields. At the foresight and innovation round table organised by IREP on 5 December 2018, Benoît Cassaigne, Executive Director at Médiamétrie, presented the company’s roadmap and major research topics which aim to keep abreast of usage trends and take full advantage of technological innovations.

 

In summary:

  • Panels and data complement one other to truly enrich audience measurement
  • Watermarking, Personal Portable Audience Measurement, a new audience meter ... media audience measurement evolves on a regular basis thanks to technological innovations
  • Médiamétrie is developing services to analyse and enrich data, providing additional information to publishers
  • Médiamétrie conducts a great deal of R&D, in particular using artificial intelligence technology: qualifying cookies, identifying logos, etc.

 

New technologies and data are enriching the panels

Technology is advancing at great speed and has a massive influence on the media behaviours of individuals, as well as the services offered and the equipment used. At Médiamétrie, innovation is guided by the evolution of practices and keeps the tempo of the customers.

Technological evolutions do not exist in opposition to the panels. On the contrary, the panels form the bedrock of audience measurement. Benoît Cassaigne reiterated, “The strength of the data produced by Médiamétrie is that it is derived from the panels.” Médiamétrie has 60,000 panellists across all its studies. “This data is deterministic, independent, durable, high quality and meticulously checked, with significant temporal depth. It also guarantees that privacy is respected.” Panels and data are complementary: data provides granularity – essential in a fragmented media world – and the panels qualify and give a reference framework for the data.

Technology presents multiple opportunities to enrich measurement and make it more powerful. A few examples: from 2008 onwards, the integration of watermarking into television audience measurement helped to overcome the technical constraints related to the evolution of reception methods. In 2011, watermarking enabled Médiamétrie to take time-shifted TV viewing via recordings into account and in 2014, it also allowed catch-up to be added. The TVM3 is another innovation, a new tablet-format audience meter that has been deployed in Médiamat panellist households since 2018. A final illustration is the pager that was developed to be worn by panellists for the particular purpose of measuring mobile audiences for Television and Radio: a Personal Portable Audience Measurement project that has undergone large-scale testing since 2018.

In 2020, Médiamat, the benchmark television audience measurement, will incorporate daily audiences for the 4 screens (TV, smartphone, tablets, computers) for all channels, thanks to data from online TV consumption (site-centric tags); the Médiamat study will also include programme viewing outside the home. According to Benoît Cassaigne, “Media consumption is increasingly individual since household sizes are decreasing, and society is becoming more individualistic. In the long term, it will make sense for media measurement to be individual. Médiamétrie’s desire is to reinforce its position as a world leader in this field.”

One major area in the development of audience measurement will be to systematically seek to enrich the panels with internet site visit data derived from tags or connection logs. This measurement hybridisation borne out of Médiamétrie’s data science and statistical expertise will increase the precision of results. These developments are made possible using cloud computing to store and process data, as well as processing methods for heterogeneous big data. The technology is also being deployed for the benefit of advertising, whether it is programmatic advertising or direct advertising. Médiamétrie is encouraging greater use of its media planning solutions in the form of APIs that can be used by ad servers. Current reflections on the development of direct advertising are going to require advertising content to be watermarked to identify the individual audience for the ad.

 

Development of Data Activities

Growing capabilities and expertise in exploiting data have led Médiamétrie to develop its data activity since 2017.

This consists of checking, enriching and qualifying the data, whether that data was collected by Médiamétrie customers through panel data or conversely, panel data that was gathered from Data supplied by publishers. It can even be both of these simultaneously via the new Data Enrichment service: it allows TV media planning to incorporate data target audiences created from the digital behaviours observed in usages of Médiamétrie’s Internet panel, or alternatively from data clients or external suppliers.

Médiamétrie’s Data Science teams have recently developed Data Profiling solutions that can call upon different methods such as deep learning and neural networks for semantic analysis and machine learning. Data Profiling allows TV and Internet publishers to associate a socio-demographic profile with their usage data, for example the composition of the household using broadband set-top boxes for TV and the age and gender of a cookie for internet. In this way, Data Profiling enables enhanced targeting of advertising. Based on a set of methods, it allows Médiamétrie to adapt to its customers’ different needs and data. Each question does require an in-depth study and a specific solution in order to be handled.

Médiamétrie is undertaking specific work to design this activity in compliance with the principles of the General Data Protection Regulation.

 

Artificial Intelligence and Blockchain present new outlooks

At its Lab’innovation, a collective of experts with complementary skills in data science, IT, and measurement technology, Médiamétrie constantly trials new solutions to prepare for the measurement of the future, including two in particular: Artificial Intelligence and Blockchain.

One current line of research focuses on using artificial intelligence to circumvent the complexity of identifying certain encrypted internet content in https: Médiamétrie is looking at an algorithm that segments mobile screenshots to extract the logo and, therefore, identify which website the mobile internet user is viewing.

Another R&D pathway focuses on pairing a video with a TV channel by identifying its logo, thanks to the creation of a specialised deep learning model using a set of image data. Research into the creation of a machine learning model to detect a radio programme type is also under way.

Ongoing reflections on Artificial Intelligence cover voice identification, sound frequencies and the measurement of the attention, engagement and emotions of individuals watching the streams.

Finally, blockchain may eventually play a major role in the use and standardisation of data. Barely 3 years old, this technology is still young. For now, its prospects remain limited by its processing speed, energy consumption and compatibility with the GDPR. Benoît Cassaigne explained, “The technology is a long way off full development, but it does present an interesting capability to manage legal and transparency issues, especially in the field of advertising. Additionally, the option to pool this data without mixing it is of particular interest to Médiamétrie.

 

Médiamétrie’s innovation policy is frequently acclaimed by interprofessional organisations abroad through some prestigious awards, including: the ICOM Grand Prize in 2015 for Hybrid TV measurement; four distinctions at the Printemps des Etudes trade fair in 2016; a double prize at the IAB Research Awards in 2017; the Grand Prize in the audience measurement category in 2018 for its Total Internet measurement; and finally, the Tony Twyman TV and Video Award at the ASI conference held in Athens in October 2018.

 

 

Laure Osmanian Molinero

3 questions for Christine Robert, Deputy Director of IREP

In your opinion, what will be the biggest challenges for research institutes in the next few years?

Research institutes are facing massive challenges resulting from the multiplicity of data collection methods, the emergence of new useable data, the diversity of information sources associated with social media and influencing, and new technologies such as AI, as well as voice, image, etc.

Challenges also arise out of a broader competitive environment that includes consulting firms, powerful technological solutions, and the internalisation of expertise which companies have developed in-house (DMP, Proprietary panels, etc.)

Against this backdrop of profound change, the issues facing research institutes relate to: recognition of their expertise (rigorous and scientific methods of collection and analysis); reassurance of their ability to transform and be open to the integration of new technologies or new approaches that are not all ‘gadget’; operational scope of the analyses and pragmatism of their recommendations.

What role do you think Artificial Intelligence can play?

Clearly, AI has a role to play for research institutes, just as it does today in many other sectors, such as health, services and/or communication, for example.

It represents a major technical advance which, when combined with the exponential growth of data, will allow institutes to perform more in-depth analysis at greater speed. AI amplifies and accelerates human intelligence and expertise, enabling it to boost its performance and be more efficient.

In summary, AI (Artificial Intelligence) serves HI (human intelligence), not the other way around!

Do you think that technologies represent the main development routes for research institutes?

Technologies are very important, but they are not the only ways research institutes can develop.

Other interesting evolutions for research institutes include: the ability to incorporate field-specific problems with more immersive working methods; the establishment of a more cross-disciplinary and less specialised organisation; the creation of partnerships with stakeholders from other spheres who possess different, specific and complementary skills and know-how.

Confidence interval calculus

Taille de l'échantillon ou d'une cible dans l'échantillon

n =

Proportion observée dans l'échantillon ou sur une cible dans l'échantillon

p =

%

Attention : ne s'applique qu'à une proportion. Le Taux Moyen est une moyenne de proportions et la Part d'audience un rapport de proportions.
Cet outil est donné à titre indicatif. Il ne saurait pouvoir s'appliquer sans autres précautions à des fins professionnelles.

Test of significance of the differences between two proportions

Permet d'évaluer si la différence entre 2 proportions est significative au seuil de 95%

Proportion

Taille de l'échantillon

Échantillon 1

%

Échantillon 2

%

Attention : ne s'applique qu'à une proportion. Le Taux Moyen est une moyenne de proportions et la Part d'audience un rapport de proportions.
Cet outil est donné à titre indicatif. Il ne saurait pouvoir s'appliquer sans autres précautions à des fins professionnelles.

En complément