Attribution methods
MMM Tradeoffs in Six Weeks
Honest tour of marketing mix modeling: priors, stability windows, and how to talk when curves disagree with intuition.
¥198,000 informational
6 weeks, 90 minutes each · Cohort
Request information
Narrative
Cohort members rebuild a toy MMM, stress-test priors, and document where the model is silent. Emphasis on reconciliation with weekly channel dashboards without forcing false precision.
Included modules
- Prior sensitivity grid with instructor feedback
- Weekly stability journal prompts
- Cross-check exercises against channel-level spikes
- Narrative drills for CFO-style questions
- Optional Julia notebook starter (no grade)
- Office hours on data sparsity for niche geos
- Template for “what we will not claim” appendix
Outcomes
- Draft an MMM limitations appendix ready for leadership.
- Choose defensible stability windows for seasonal categories.
- Explain two reconciliation paths between MMM and weekly ops metrics.
Responsible mentor
Jonas Mehta
Attribution mentor focused on Bayesian workflows for distributed teams.
Participant questions
No. You can stay entirely in prepared notebooks; engineers may fork deeper if they choose.
Experience notes
MMM Tradeoffs made our priors debate productive instead of political. Still disagree on TV decay, but we now log why.
Anonymous cohort member: the CFO narrative drills were uncomfortable—in a useful way.
★★★★☆ Verified learner — loved the stability journal, wanted one more retail-specific example.