Agentic Marketing Science: MMM, Attribution and Experiments with Claude Code
Gui Diaz-Berrio
Marketing Analytics Leader & Author
You understand marketing science. The code is where it falls apart.
Marketing measurement is changing fast. Google Meridian just launched. Bayesian MMM is replacing legacy approaches. Your team is asking for attribution answers you can't deliver with last-click.
But the gap between knowing MMM theory and running it is brutal. 54-column datasets where 12 variables will break your model. Thousands of adstock parameter combinations. Meridian errors the docs don't cover. Model outputs you're not sure you can trust.
This course closes that gap — with AI doing the heavy lifting.
You'll build MMM models (OLS, Bayesian, and Google Meridian), run Shapley and Markov attribution, and design geo-experiments with power analysis. Not toy examples — real pipelines on real data.
The difference: you leave with Claude Code Skills — AI-powered tools that automate variable classification, data formatting, and model diagnostics. Run them on your own data, at your job, after the course ends.
You don't need to be a Python expert. The Skills handle implementation. You focus on the science and the judgment calls.
Early bird: use code EARLYBIRD-MAY26 for 30% off ($450 savings).
What you’ll learn
Go from understanding marketing measurement to running it at work — with AI-powered tools you keep forever.
Apply geometric adstock and Hill saturation transforms to model carry-over effects and diminishing returns across channels
Build and diagnose an OLS model with VIF checks, p-values, Durbin-Watson tests, and coefficient sign validation
Construct a Bayesian MMM with PyMC — set priors, run MCMC sampling, and quantify uncertainty around channel ROI
Prepare raw data for Meridian: dimension naming, NaN handling, VIF checks, and the formatting that trips most users
Estimate media effects across geographies and interpret Meridian's contribution and ROI outputs
Use the `meridian-model` Claude Code Skill to automate data prep and catch silent formatting errors
Build Shapley and Markov attribution models on multi-touch journey data to replace last-click with causal allocation
Compare attribution outputs across methods and explain why they disagree — and which to trust
Connect attribution insights to MMM results for a unified view of channel performance
Run power analysis to determine required sample size, duration, and detectable effect before spending budget
Match treatment and control regions, run difference-in-differences estimation, and validate experiment results
Feed experiment results back into your MMM as calibrated priors — closing the measurement loop
Use purpose-built AI Skills for variable classification, model diagnostics, and data formatting — not generic ChatGPT prompts
Run the full measurement pipeline on your company's data during the Week 3 BYOD workshop
Keep and adapt the Skills after the course — they encode the methodology, so you can repeat the analysis independently
Learn directly from Gui
Gui Diaz-Berrio
Led measurement for €100M+ in ad spend at Kindred Group & Just Eat Takeaway
Who this course is for
Marketing analyst or data scientist who's been tasked with building in-house measurement but struggles to go from theory to working code
Marketing lead who relies on MMM vendor's black box and wants to understand, validate, and eventually complement or replace their outputs
Data professional who tried Meridian, Robyn, or PyMC-Marketing and hit walls with data formatting, prior specification or debugging errors
Prerequisites
Working knowledge of Python
You can write a for loop, define a function, and load a CSV with pandas. You do NOT need to be a Python expert — AI handles the heavy code
Familiarity with marketing or media data
You've worked with campaign data, channel spend, or marketing KPIs. You know what impressions, CPM, and ROAS mean in context
What's included
Live sessions
Learn directly from Gui Diaz-Berrio in a real-time, interactive format.
Lifetime access
All notebooks, datasets, recordings, and readings — yours forever, not just during the cohort
Discord community
Get help between sessions, share your work, and connect with students from all cohorts — not just yours
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Claude Code Skills you keep
AI-powered tools for variable classification, Meridian data prep, and model diagnostics. run them on your own data after the course
Production-ready Python utils package
Adstock transforms, saturation curves, OLS pipeline, contribution analysis, and geo-experiment helpers — ready to use at work.
Real datasets, not toy examples
Multi-geo synthetic data modeled on real advertising patterns, plus 10K multi-touch journeys for attribution exercises
BYOD workshop
Bring your own company data in Week 3 and start adapting the pipeline to your real use case
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Course syllabus
53 lessons • 6 projects
Week 1
Course Intro
Week 1: The Measurement Landscape
Capstone Project
Module 1: Statistics Fundamentals
Module 2: MMM Foundations
Module 3: Advanced issues when building your MMM
Free resources
Schedule
Live sessions
4-6 hrs / week
Practical Workshops to work on your own data
Projects
2 hrs / week
Async content
2-6 hrs / week
Fundamental Theory and Building Blocks to be able to advance to practice
Frequently asked questions
$1,500
USD