Blind Spots in Digital vs. Traditional Media Measurement
Laerte Finotelli analyzes the limits of walled gardens, the illusion of deterministic data, and the rise of attention metrics.
In a discussion on Marketing - The Podcast, Laerte Finotelli examines the challenges of unifying measurement across closed digital channels and traditional media. The conversation explores the differences between probabilistic sampling and device counting, alongside the real hurdles in evaluating attention.
Laerte Finotelli, product coordinator, discussed the challenges of integrating data across diverse media channels and the limitations of current digital metrics on Marketing - The Podcast. In a conversation with Bruno Gimenes, Finotelli analyzed how the isolation of major platforms and technical counting assumptions impact decision-making for brands and advertisers.
The video above features the full discussion on data science applied to media products, including practical examples of research methodologies and user behavior across multiple screens.
The limits of walled gardens in measurement
The rise of so-called walled gardens—closed ecosystems owned by major tech companies that do not share raw data with external markets—creates barriers to calculating unique campaign reach across media. Finotelli emphasizes that this lack of interoperability makes it impossible to determine whether the same individual was reached across digital channels and traditional mediums like TV, radio, print, or out-of-home.
"Looking at the ecosystem, at the industry as a whole, for agencies as well as brands and advertisers, greater collaboration would be extremely valuable," Finotelli explains. In his view, tools do not necessarily need to adopt identical indicators, but they must converge to enable true reach deduplication rather than merely tallying isolated views.
Probabilistic versus deterministic data
Comparing traditional research methods with digital reporting involves fundamentally distinct methodological perspectives. The historical television model, such as IBOPE, relies on probabilistic sampling of households and individuals to project total audience within a population universe. Conversely, digital platforms typically position themselves as deterministic because they record the exact event triggered on a server.
Finotelli points out that this technical precision masks significant blind spots regarding who is actually consuming the content: "Then you ask: 'Who is this person?' 'Oh no, I don't measure people, I measure devices. I measure logged-in accounts.'" Without accurate demographic identification of the individual in front of the screen, deterministic device counts fail to address advertisers' need to evaluate audience profiles against a unified baseline.
The fragility of engagement metrics
The pursuit of qualitative metrics has led the industry to explore attention standards as an alternative to purely quantitative indicators like impressions and GRPs. Finotelli critically assesses the conventions adopted by video platforms to record consumption and interaction.
When commenting on how online platforms classify views, Finotelli highlights the gap between technical definitions and actual retention: "They created a new metric called engaged views, which requires just 2 seconds of video pixels on the screen. Calling something engaged after just 2 seconds feels dangerous to me."
Context, exposure quality, and attention models
The effectiveness of an advertising message relies on contextual factors that quantitative reports fail to capture, such as immersion level, enabled audio, and the proportion of the screen occupied by the creative. A fleeting glance in a social media feed does not carry the same impact as an ad displayed in a movie theater.
Finotelli points to experiments conducted in the UK using mobile screen tracking where a user's thumb covered the ad while scrolling, even though the platform registered the impression as delivered. To address these discrepancies, emerging methodologies combine brand recall studies, eye tracking, and artificial intelligence models to evaluate the true value of each touchpoint.