Why New As-Run Data Could Reset Radio's MMM Role

RADIO ONLINE | Monday, February 2, 2026 | 2:27pm CT

Cumulus Media and Westwood One

A new blog post from Westwood One argues that media mix modelers should reassess how radio is evaluated, citing the availability of weekly, as-run radio delivery data as a meaningful "trend break" from historical measurement practices.

The post, published by Cumulus Media | Westwood One's Audio Active Group and authored by media analytics veteran John Fix, contends that many media mix models (MMMs) still rely on historical guardrails that constrain radio's return on investment to past performance-an approach he says is no longer appropriate given improvements in data quality and granularity.

Fix notes that traditional MMMs often depend on planned media inputs, which can create a smoothing effect that masks real-world delivery variation and weakens the ability to correlate AM/FM radio advertising with sales. In contrast, newly available as-run radio data captures actual weekly delivery, producing a stronger signal for modeling and attribution.

The blog points to recent industry developments that have expanded access to this data, including broadcaster partnerships with Media Monitors, Act1, and Nielsen to formalize as-run radio measurement. Advertisers can now obtain weekly, delivered GRPs at the DMA level, addressing long-standing concerns that radio data lacked sufficient detail for effective modeling.

Momentum increased further in 2025 when Media Monitors expanded its radio monitoring footprint from 106 to 250 U.S. markets, significantly improving national coverage. The post also references guidance shared in 2024 by Dave Hohman, EVP and GM of Global Marketing Mix at Circana, who emphasized the importance of using as-run data, DMA-level delivery, and adequate GRPs when incorporating radio into MMMs.

Given these changes, the blog recommends a full reset in how radio is modeled. Among the key takeaways: modelers should ensure the entire radio dataset is based on as-run delivery rather than appending new data to legacy models; they should acknowledge that current radio data differs fundamentally from historical inputs; and they should consider treating radio as a "new" media channel when establishing benchmarks and priors.

Fix also addresses the role of Bayesian priors in MMM, suggesting that reliance on tight, historically informed priors may suppress the insights available from improved radio data. Instead, he advocates for weaker or flatter priors that allow models to explore a wider range of outcomes, particularly when using datasets that have rarely-or never-been modeled before.

The post concludes that MMM should remain a collaborative process among advertisers, broadcasters, and modeling firms, with greater transparency around data inputs, assumptions, and the influence of historical norms-especially as new measurement methodologies reshape how radio performance is evaluated.

Read the entire blog post here.