What we're reading.
Marketing Mix Modelling (MMM) and measurement reading from around the industry, curated weekly and sorted by kind: what's new, how to do it, what you can now measure, where to learn it, and what people think. Each entry links to the source, with one line on why it's worth your time.
News
Launches, releases and new industry research.
1- New IPA and SEF report identifies the drivers of sports sponsorship effectiveness
Identifies drivers for sports sponsorship effectiveness – useful for defining the right variables and lag structures when modelling long-term brand impact.
More on this The four disciplines of Marketing Mix Modelling. · our writing
IPA
Techniques
How to model, test and read the results.
11- Planning a Marketing Mix Modelling Project
Links technical model choices to business beliefs – arguing that stakeholder alignment is more critical for project success than the underlying econometrics.
More on this What an adstock is, and why the rate matters. · our writing
James Walker on LinkedIn 15 Oct 2026 - The Budget Was Never Randomized: What a Marketing Mix Model Can and Cannot Tell You
Highlights how non-randomised budget allocation biases model results – essential reading for anyone using MMM to justify future spend levels.
More on this OLS and Bayesian estimation, explained. · our writing
Marketing Mix Modeling on Medium 2 Oct 2026 - I Pitted XGBoost Against Logistic Regression on 358 Matches. The Boring Model Won.
The data comes from sports betting, but a simple model beating XGBoost on 358 matches is a useful parallel for the small datasets typical of marketing measurement.
More on this OLS and Bayesian estimation, explained. · our writing
Towards Data Science 28 Jun 2026 - When MMM and Incrementality Disagree: What Is Each Actually Telling Us?
Details why model estimates and experiment results often diverge – a vendor-written but technical guide to using lift tests as Bayesian priors.
More on this Papilon 0.3.0: · TechniquesCounterfactuals: A Practitioner’s Guide · Techniques The last click isn't marketing. · our writing
Marketing Mix Modeling on Medium 11 Oct 2026 - I Hid Four Traps in a Forecasting Task. Here Is What Four AI Assistants Did.
LLMs fail to spot data leakage and structural breaks in forecasting tasks – a warning for those automating measurement workflows.
More on this The Great MMM Comeback Is Half a Lie · PerspectivesA Google Team Measured Half of My Argument, and Left the Other Half Open · Perspectives The four disciplines of Marketing Mix Modelling. · our writing
Towards Data Science 6 Oct 2026 - Why Powerful ML Is Deceptively Easy — Part 2
This technical examination of data leakage explains why model accuracy is often overstated, offering practitioners specific criteria to challenge the validity of results derived from complex machine learning architectures.
More on this A Google Team Measured Half of My Argument, and Left the Other Half Open · PerspectivesI Hid Four Traps in a Forecasting Task. Here Is What Four AI Assistants Did. · Techniques The marketing ROI that was actually a price cut. · our writing
Towards Data Science 1 Jul 2026 - How Wrong Is Your Marketing Mix Model (MMM)?
Static media plans often provide insufficient data for accurate measurement – changing flighting patterns can improve the signal and provide more reliable results.
More on this OLS and Bayesian estimation, explained. · our writing
Towards Data Science 7 Oct 2026 - Papilon 0.3.0:
Benchmarks a Bayesian model against known ground truth to show where it fails; written by the developer, but the performance data is honest.
More on this The Great MMM Comeback Is Half a Lie · PerspectivesWhen MMM and Incrementality Disagree: What Is Each Actually Telling Us? · Techniques OLS and Bayesian estimation, explained. · our writing
Marketing Mix Modeling on Medium 24 Oct 2026 - Counterfactuals: A Practitioner’s Guide
A rigorous technical breakdown of counterfactual construction and propensity score matching, with code for applying it in incrementality testing. Written by a software vendor.
More on this When MMM and Incrementality Disagree: What Is Each Actually Telling Us? · Techniques The last click isn't marketing. · our writing
Recast 15 Jun 2026 - More Controls Isn’t Always Better: Choosing the Right Ones for Your Business
A useful technical reminder that adding more control variables can introduce bias rather than remove it – a sound basis for challenging a model specification. Written by a software vendor.
More on this OLS and Bayesian estimation, explained. · our writing
Recast 22 Jun 2026 - The Model Was Right. The Data Was Not.
This technical walkthrough demonstrates how omitting price and promotion variables leads to inflated marketing return on investment (ROI), offering a clear case for why data quality often matters more than model architecture.
More on this A Google Team Measured Half of My Argument, and Left the Other Half Open · PerspectivesI Hid Four Traps in a Forecasting Task. Here Is What Four AI Assistants Did. · Techniques The marketing ROI that was actually a price cut. · our writing
Medium · MMM tag 29 Jun 2026
Perspectives
Opinion and commentary worth your time.
4- A Google Team Measured Half of My Argument, and Left the Other Half Open
Argues that technical performance metrics are secondary to a clear business brief – theoretical, but a useful reminder to define success before commissioning analysis.
More on this I Hid Four Traps in a Forecasting Task. Here Is What Four AI Assistants Did. · TechniquesWhy Powerful ML Is Deceptively Easy — Part 2 · Techniques The marketing ROI that was actually a price cut. · our writing
6 Oct 2026 - Binet: Think big like investors to avoid the returns death spiral
This summary of Les Binet’s recent talk provides clear arguments for practitioners needing to justify long-term brand investment to finance partners focused on immediate attribution.
More on this The four disciplines of Marketing Mix Modelling. · our writing
22 Jun 2026 - The Great MMM Comeback Is Half a Lie
An assessment of the risks in automated, black-box modelling tools that put speed ahead of causal rigour – worth reading before your next vendor demo.
More on this Papilon 0.3.0: · TechniquesI Hid Four Traps in a Forecasting Task. Here Is What Four AI Assistants Did. · Techniques Three questions before you sign. · our writing
20 Jun 2026 - (Almost) Everybody Hates MMMs
Worth reading for its account of why models so often fail to move budgets: what the analysis delivers rarely matches what executives need to decide.
More on this The Great MMM Comeback Is Half a Lie · Perspectives Three questions before you sign. · our writing
11 Jun 2026
Nothing matches yet. It fills up weekly – or try searching the whole site.