Observed Signal · Oct 6, 2026 · Podcast Episode · Source: AdExchanger · Impact: 2/5 · Sentiment: Neutral
AI-Built MMM Needs Thesis for Trust
In a recent episode of AdExchanger's Inside the Stack podcast, Madan Bharadwaj, Founder and CEO of M-Squared, discusses the challenges and methodologies behind AI-built marketing mix models (MMM). Bharadwaj highlights that while AI has reduced model construction time from months to days, the need for expert judgment has increased. He advocates for triangulation, running MMM and incrementality tests in parallel to validate results, and stresses the importance of starting with a thesis about go-to-market strategy to guide measurement. The episode also covers a case study of a beauty brand where AI recommended cutting $30 million from Amazon ad spend, but SKU-level modeling revealed different opportunities, leading to 5-7% top-line growth. Bharadwaj also touches on the need for finance to understand marketing results and the future of measurement moving towards 'super cognition'.
The article provides expert insights on AI-driven measurement practices but does not announce a major industry event or product launch.
Track Real-Time Measurement Signals & Market Shifts
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Madan Bharadwaj is Founder and CEO of M-Squared.
- AI has reduced MMM build time from months to days.
- Bharadwaj recommends triangulating MMM with incrementality tests.
- A beauty brand case study showed AI suggested cutting $30M from Amazon ads, but SKU-level modeling found 5-7% top-line growth.
- Measurement is heading toward 'super cognition' according to Bharadwaj.
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
AI in Marketing: Hype vs. Reality in Measurement
AI, particularly LLMs, has generated hype around marketing measurement but can deliver wrong, unvalidated analyses in media mix modeling (MMM). LLMs are not designed for causal inference, so attribution results from such models can mislead budget decisions despite confident outputs. Vendors marketing AI-powered MMM may obfuscate methodologies and skip essential validation, increasing risk for enterprise marketers. AI can accelerate model estimation and support peripheral tasks—summarizing outputs, clarifying assumptions, and flagging data anomalies—but it cannot replace core causal modeling or guarantee profit. The article urges using an independent validation framework that emphasizes experimentation, forecast reconciliation, and model-quality checks to confirm true causal signals. Alaska Airlines is cited as adopting an open-source MMM solution; with a nine-figure marketing budget, small forecast errors can cause multimillion-dollar misallocations if tooling is unvalidated. Overall, substantial progress comes from demonstrable ROI rather than AI magic, and firms should demand evidence of profit impact over hype.
On Device Research Launches Brand Lift Data API
On Device Research, a mobile research company, has launched a Brand Lift Data API designed to measure brand lift in the 'outcomes era'. The API enables advertisers to integrate brand lift measurement directly into their data workflows, moving beyond traditional survey-based methods. It provides real-time, privacy-compliant brand lift insights by leveraging device data. The launch reflects a broader industry trend towards more direct, data-driven measurement solutions that align with performance marketing goals, offering a more agile and granular approach to understanding advertising effectiveness.
WBD White Paper Calls for Measuring TV Ad Basics
Warner Bros. Discovery (WBD) published a white paper titled 'The Unified Reach Paradox' during Advertising Week New York, advocating for standardized TV ad measurement fundamentals. The paper presents findings from WBD's investigation of five major measurement providers—VideoAmp, iSpot, Comscore, Innovid, and Samba—across more than 50 campaigns on linear and streaming. WBD found significant variance in universal audience estimates, impression counting, and cross-platform coverage. For example, Comscore counted 43.5M impressions for a campaign while iSpot counted 24.6M. A survey of 154 agency and marketer decision-makers revealed that 73% consider unified reach a critical missing metric, and 79% cannot fully trust outcomes without it. WBD calls for industry standards on impression and viewability counting, and transparency in projection methodologies, urging a 'nutrition label' approach.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.