MMM optimizers only exploit: bandit-style exploration recovers lost revenue
twiecki · x · 2026-10-07
Pymc Labs argues that typical Marketing Mix Model (MMM) optimizers only exploit, so their channel response curves stop learning and revenue is left on the table.
Their fix: treat ad channels as bandit arms and add aimed exploration on top of pure optimization. In practice this surfaced an underrated channel and made more money than pure optimization alone.
Key takeaway: MMM optimization needs ongoing learning about channel curves, not just current-best budget allocation.
More from Research
- Scale AI's Muse Claims 6 Open Math Problems Solved With Mathematicians — inductionheads · 2026-10-07
- "The Hodge Conjecture Has Fallen": Unverified Claim of AI Math Breakthrough — rand_longevity · 2026-10-07
- Uni-LaDiR unifies reasoning across modalities via latent diffusion thoughts — Lianhuiq · 2026-10-07
- Kakeya Conjecture in R^3 Resolved by Hong Wang, Building on Her Fields Medal Work — teortaxesTex · 2026-10-07
- Research note: filtering subversion-related info from pretraining is feasible — jammastergirish · 2026-10-07
- Mathematician admits OAI's Hodge conjecture progress outpaced expectations, mocks the hype — ctjlewis · 2026-10-07