google/meridian

What it solves

Meridian addresses the challenge of measuring the impact of marketing campaigns on revenue or key performance indicators (KPIs) without relying on user-level cookies or private data. It helps advertisers determine their return on investment (ROI) and optimize future budget allocations across various marketing channels.

How it works

Meridian is a Bayesian causal inference framework that uses aggregated data to perform Marketing Mix Modeling (MMM). It employs a holistic MCMC sampling approach called the No U Turn Sampler (NUTS) and leverages GPU acceleration via tensors to reduce training time and provide real-time optimization results.

Who it’s for

It is designed for advertisers and marketing analysts who want to run in-house models to analyze media effectiveness and plan budgets based on geo-level or national-level data.

Highlights

• Privacy-safe: Uses aggregated data instead of user-level information.
• Customizable: A highly flexible modeling framework based on Bayesian causal inference.
• Calibration: Supports calibration using experiments and prior information.
• Optimization: Able to optimize target ad frequency using reach and frequency data.
• Scalable: Handles large-scale geo-level data and national-level modeling.