OptiMix uses Automatic Differentiation Variational Inference (ADVI) — a modern approximate Bayesian inference algorithm — to estimate marketing channel effects. Unlike MCMC methods that take hours or days, ADVI runs in minutes while producing full posterior distributions over every channel contribution.
The Media Mix Model
The core model is a hierarchical Bayesian regression accounting for: Baseline (Organic) sales, Saturation (diminishing returns at higher spend), Adstock (carry-over decay), Trend & Seasonality, and Control Variables (price, promotions, competitor activity).
Privacy-First
OptiMix runs on aggregated weekly spend and revenue data — no PII, no cookies, no third-party tracking. GDPR and CCPA compliant out of the box.
Minimum Data Requirements
26 weeks of weekly spend and revenue data across two or more marketing channels. 52+ weeks enables seasonal pattern detection. 2+ years separates long-term trend from media impact.
What OptiMix Does Not Do
It does not replace incrementality testing, does not provide real-time attribution, does not optimize creative or audience targeting, and does not manage automated bids. See the pricing page to get started.