Nielsen wants to banish batch learning with new AI system

Audience measurement giant Nielsen is the latest to jump on the artificial intelligence bandwagon, releasing its new Nielsen AI product plug-in for its Marketing Cloud clients this week – claiming to use adaptive learning technology to automate audience optimization and make “real time” responses to changes in consumer behavior.

The grey area of AI and machine learning technologies continues to throw up confusion and contradiction even for industry professionals and, most importantly, investors. As much as we like to poke fun at Nielsen’s failings, a company with as much data clout and industry experience as it has, means Nielsen should produce a rather reliable and effective piece of artificial intelligence technology.

Nielsen AI says its audience optimization functionality works by monitoring streams of device-linked data, such as motivations, interests and actions – data pulled from web or app usage history, e-commerce purchases, or Nielsen’s own audience data.

Marketers can then adapt campaigns to match any shifts in consumer media and buying behavior, including market dynamics such as seasonal and local market demand, competitive actions and advertising, as well as the thousands of consumer attributes. Nielsen AI says it can amalgamate audience data updates across platforms including OTT video sites and apps, social media, email, programmatic, and search.

When Nielsen came out with its total content ratings (TCR) system, this seemed like a promising solution to Nielsen’s struggles to adapt its existing ratings system in order to track cross platform viewing, but this quickly fell out of favor with networks such as NBC, which said it wasn’t ready for primetime. Although in its infancy, Nielsen AI could be a step towards achieving total audience data, and time will tell if the major networks sentence it to the same fate as its premature TCR system.

The wealth of data at the hands of the likes of Nielsen is of course set to get even deeper and more insightful – thanks to President Trump’s signature finally landing on the bill to allow ISPs to sell on sensitive data such as location, web history and financial information to advertisers without user permission (see separate story in this issue).

Instead of saying it uses machine learning, Nielsen AI has instead opted for “adaptive learning technologies”. We assume these are some budding machine learnings algorithms in the making – perhaps not wanting to peg itself as a fully-fledged provider of machine learning capabilities just yet, as many companies do too hastily.

However, what the firm does note is that Nielsen AI’s adaptive learning is a step away from batch (or offline) learning, a more widely used method of data analysis in the industry. Nielsen claims that a batch model performs the majority of its analysis overnight, and therefore is too late to the party to adapt in time to deliver the appropriate audience experiences.

Nielsen’s data-as-a-service version of its Marketing Cloud application will also see the inclusion of Nielsen AI. It has over 60,000 audience segments, allowing advertisers to custom build campaigns based on specific consumer interests and behaviors. It also claims to be able to “dynamically hypertarget viewers” – the space-age equivalent of personalized ads is already with us, given Nielsen’s take on things.

Networks are upping the stakes to give marketers more control over data-driven advertising on linear TV, with NBCUniversal recently announcing it will open up access to its Audience Targeting Platform (ATP) and its programmatic offering NBCUx.

Nielsen’s demographic data is no longer deep enough in today’s demanding targeted advertising landscape, so NBCUniversal said it is sidelining $1 billion worth of ad inventory to be sold on metrics that drill down further into the consumer experience than Nielsen can provide. The figure is around one sixth of NBCU’s total ad sales.

CBS also came out last month, saying it will be giving out total content ratings from Nielsen which measure DVR playback and VoD content on top of traditional linear viewing, and claims it will add online and mobile viewing in the future.

EVP of Nielsen’s Marketing Cloud arm, Mark Zagorski, commented, “Nielsen AI takes a big leap beyond industry-standard batch-learning approaches, which are limited by static audience data sets and learning processes that can take days to complete. In a competitive environment where every moment counts, marketers need to be able to act on up-to-the-second information in an automated way. Nielsen AI equips them with the tools they need for real time data processing, learning and syndication, enabling marketers to cut through the clutter by providing superior customer experiences across channels, devices and time.”99