The leader track 🤝 · for CMOs, CDAIOs & boards · 6 x 45 min · no code
Attribution is a credit-splitting rule you choose, not a fact you measure - and the rule quietly moves your whole budget. This track teaches you to read the models, spot what each one flatters, and defend a spend reallocation to the board. Also a friendly on-ramp for LinkedIn readers.
The credit problem: who gets the sale?
One real journey, eight models, eight different answers - and the reason your last-click dashboard has been quietly lying to you.
📐 Leader session 2The heuristics you already use
First, last, linear, time-decay, position - what each rule biases toward, and how to read a platform report critically.
🧠 Leader session 3Data-driven attribution, decoded
Markov's removal effect and Shapley's fair share - explained without math - plus why GA4's data-driven model is a black box.
⚖️ Leader session 4MMM vs MTA vs incrementality
The same question answered three ways, why the numbers disagree, and which method to trust at your budget tier.
🍪 Leader session 5The 2026 cookieless break
How ATT, SKAN, and cookie loss broke multi-touch, why mix modeling and lift tests came back, and what to demand from your team.
💰 Leader session 6From credit to budget
Turn attribution output into a spend reallocation a CDAIO signs - triangulate, move at the margin, and confirm it with a holdout test.
The builder track 🛠️ · for practitioners · 10 x 45 min · SQL + Python
For DA, DE, DS, and growth engineers. One running brand - Lumen Skincare - measured every way: build the journey table in SQL, then implement heuristics, Markov, Shapley, ML/DL attribution, GA4 DDA reconciliation, marketing mix modeling, and causal geo-lift tests - from scratch, on the same data.
The touchpoint data model + SQL journeys
The 3-table data model and the window-function SQL that turns raw events into ordered customer journeys. The step no course teaches.
📐 Builder session 2Heuristics in SQL and Python
First, last, linear, time-decay, position - implemented on Lumen, reconciled into one table where the same journey gets six answers.
🪜 Builder session 3Why heuristics fail, the data-driven leap
The three failures of fixed rules, then Shao-Li's bagged logistic regression - the historical bridge from heuristics to learned credit.
🔗 Builder session 4Markov attribution from scratch
Transition matrices and the removal effect, worked by hand then in numpy - plus why you must normalize and when to go higher-order.
🎲 Builder session 5Shapley value attribution from scratch
Coalitions, marginal contribution, and the efficiency axiom - exact Shapley with itertools, Monte-Carlo for scale, and the SHAP bridge.
🤖 Builder session 6ML and deep-learning attribution
Gradient boosting + SHAP as the production path, LSTM-and-attention models (DNAMTA, DeepMTA), and when deep learning just overfits.
🔬 Builder session 7GA4 DDA + reconciling tools
What Google's data-driven model actually does, and how to de-duplicate the same conversion that GA4, Meta, and TikTok all claim.
📈 Builder session 8MMM: adstock and saturation
Marketing mix modeling on aggregate spend - carryover, Hill saturation curves, and Robyn vs Meridian vs PyMC-Marketing.
🧪 Builder session 9Incrementality and geo-lift
The only causal leg: holdout design, synthetic control, and feeding measured lift back to calibrate your MMM priors.
🏛️ Builder session 10The unified 2026 stack
Assemble the calibrated loop - MMM anchor, lift-test calibration, DDA tactics - plus the eleven pitfalls and the ethics of modeling on incomplete data.
Choose your path 🗺️
Leaders start at Leader session 1; practitioners start at Builder session 1. The fast-tracks fit a specific gap.
The knowledge map 🧠
The builder track at a glance - hover a session to spotlight its concepts, click any node to jump in.