Hershey partners with Mutinex and Tracer to replace slow, manual MMM cycles with an agentic AI system running monthly across its full brand portfolio.
Key Takeaways
Previously, Hershey ran MMM three times a year for ~5 brands; results arrived months late, e.g. full 2024 data read arrived mid-2025.
Mutinex uses a multi-agent architecture (Claude + Gemini) with specialist agents for econometrics, competitive pricing, and model failure diagnosis.
Tracer handles data cleaning and standardization across fragmented marketing and retail systems, enabling Mutinex models to run in ~3 weeks.
Hershey is moving to monthly MMM across its full portfolio, targeting 4-5% increase in revenue attributable to media spend.
The $2B+ in combined media and trade spend can now be evaluated and reallocated on a monthly rather than annual basis.
Hacker News Comment Review
Commenters largely dismissed the AI angle as a distraction from product quality issues, with several pointing to Hershey’s chocolate degradation over time as the actual business problem.
The one technical signal commenters acknowledged: Tracer’s CCO quote that “most companies don’t have an AI problem, they have a data readiness problem” – commenters read this as an admission that the AI layer is secondary to basic data infrastructure.
Notable Comments
@kotaKat: Reads the data-readiness quote as companies using AI to “hallucinate the data they wanted to imagine” rather than face real customer signals.
