← All terms·Cross-media measurement
Reference library · Cross-media measurement
Panel-Based vs Big Data Measurement
Panel-based measurement projects audience behavior from a small, recruited sample of people or households whose exposure is tracked (people meters, diaries) and weighted to represent the total population. Big-data measurement counts actual exposures at or near census scale, from set-top boxes, smart TVs, or platform server logs, without sample-to-population extrapolation. Panels deliver a stable, demographically verified currency comparable across media owners; big data delivers volume and granularity but usually covers only its own platform's footprint and lacks verified person-level demographics.
Worked example
A prime-time spot on a Polish broadcaster is measured two ways. The Nielsen panel (2,700 households projecting a universe of 13.9 million TV households) records 43 panelists in the target 25-49 group tuned in during that minute, projecting to a 2.4% rating and roughly 379,000 contacts in a 15.8 million target universe. The same spot's smart-TV census data, covering 3.1 million connected sets, logs 612,000 sets tuned to that channel-minute, a 19.7% share of connected devices. The two numbers aren't directly comparable: the panel figure is a population-projected, demographically verified contact count, while the big-data figure is a raw device count skewed toward younger, urban, connected households, with no guarantee anyone was watching.
How it is used
Panel data counts a small number of recruited, weighted individuals minute-by-minute and expresses results as rating and reach against a defined population universe; big-data sources count devices, sessions, or server-logged impressions against whatever footprint that platform actually covers, with no weighting to the wider population. Planners use panel currency to compare TV against TV and, increasingly, against digital in a single cross-media metric, and use big data to sanity-check delivery at scale, catch under-delivery, or fill gaps panels can't reach (long tail channels, small target cells). The recurring mistake is treating a big-data percentage or count as if it were a panel rating: it isn't demographically verified, isn't projected to a total population, and its coverage is limited to opted-in or connected devices, so it systematically over- or under-states audience depending on who owns those devices.
The common mistake
Don't quote panel ratings and big-data device counts side by side as if they were the same currency; state which population each figure is projected against before comparing them.