An opinionated playbook for data science leaders: connect brand tracking, Bayesian MMM, experiments, and VARX to commercial decisions.
Read the playbookWriting
On AI, business decisions, and building teams.
Archive
The Agentic Data Science Playbook
A practical playbook for framing investigations, equipping data science agents, reviewing their results, and turning lessons into reusable expertise.
Sep 30, 2026 · Published elsewhere ↗
How to build a brand measurement system: a Bayesian playbook
An opinionated playbook for data science leaders: connect brand tracking, Bayesian MMM, experiments, and VARX to commercial decisions.
Sep 15, 2026
The 25 Most Asked Questions About Bayesian MMMs
Practical answers for marketing leaders and data scientists on building, validating, interpreting, and acting on Bayesian marketing mix models.
Aug 4, 2026
How I Verified Climate Change While Planning a Trip to Berlin
A Bayesian detour through 66 years of rain data. The model couldn’t beat the naive baseline, and proved that nothing can. The trend coefficient was the real payoff.
Jul 3, 2026
No One Said the Ferrari Luce Felt Like a Ferrari
I ran 850 synthetic interviews to test Ferrari’s electric Luce, showing how AI panels can speed concept, campaign, and website research.
Jun 16, 2026
Which MMM Priors Actually Matter: A Sensitivity Analysis
A sensitivity analysis across 14 prior configurations reveals which Bayesian MMM priors are inert, which remain influential, and where ROAS error persists.
May 10, 2026
What Happens When Stakeholders Can Talk to Bayesian Models?
A working agent on top of a PyMC-Marketing model shows how stakeholders can question Bayesian MMMs in plain English without routing requests through a data scientist.
May 5, 2026
Bridging Tomorrow’s Uncertainty: Combining Chronos2 Foundational Models with PyMC-Marketing for Marketing Forecasting
How to combine foundational time series models with Bayesian Marketing Mix Models for scenario planning when future control variables do not exist yet.
Mar 17, 2026 · Published elsewhere ↗
What Do Data Teams Actually Look Like in 2026? A Large LinkedIn Benchmark
I analyzed LinkedIn data on 213 tech companies to benchmark data and AI team composition. Role mix matters more than team size.
Mar 14, 2026
I ran 25 experiments on AI brainstorming. More agents beat better prompts.
Experiments across 1 to 22 AI agents show why scaling agent diversity can matter more than better prompts for generating varied ideas.
Mar 14, 2026
Bayesian Autoresearch for Causal Inference: When You Can’t Score the Thing You Care About
Adapting Karpathy’s autoresearch to Bayesian causal modeling, where the treatment effect has no ground truth and the agent must optimize a proxy.
Mar 11, 2026
I Gave 24 LLMs a Personality Test. Their Answers Say More About Training Than You’d Expect.
A Big Five survey across 24 LLMs finds distinct behavioral fingerprints by model family—and shows why the statistical patterns need careful interpretation.
Feb 23, 2026
100,000 Agent Skills in 3 Months. Most of Them Are Untested.
The AI agent skills ecosystem grew explosively. Security scanning followed. But nobody is checking whether skills actually work. That’s the gap.
Feb 22, 2026
Agentic Data Science: Three Pillars of Trustworthy AI
Why shipping judgment is harder than shipping code, and three pillars for building AI agents you can trust.
Feb 18, 2026 · Published elsewhere ↗
Let Bayes tune Bayes: hyperparameter optimization for causal MMMs with Optuna
Combining Optuna’s Bayesian optimization with PyMC-Marketing and CRPS to systematically tune Media Mix Model hyperparameters, instead of guessing.
Dec 20, 2025
Introduction to Bayesian Thinking for Business Leaders
Why traditional statistical methods often fail in business contexts, and how Bayesian approaches can help you make better decisions under uncertainty.
Nov 12, 2025
Synthetic Consumers: The Promise, The Reality, and The Future
How LLMs are changing market research and product testing through simulated consumer behavior.
Nov 1, 2025 · Published elsewhere ↗
LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings
A method enabling LLMs to generate realistic consumer survey responses, achieving 90% of human test-retest reliability.
Oct 9, 2025 · Published elsewhere ↗
AI Innovation Lab: Reimagining Product Innovation with Synthetic Consumers and Agentic Workflows
An end-to-end platform combining agentic workflows and synthetic consumers for every stage of product development.
Jun 3, 2025 · Published elsewhere ↗
Turning Customer Feedback into Action: An LLM Blueprint
A systematic approach to analyzing app reviews at scale using large language models.
Dec 1, 2024 · Published elsewhere ↗
Rethinking Customer Lifetime Value using Machine Learning at HelloFresh
Building Morpheus, a machine learning algorithm for weekly customer-level lifetime value predictions.
Dec 7, 2020 · Published elsewhere ↗
Bayesian Media Mix Modeling using PyMC3, for Fun and Profit
How HelloFresh built a Bayesian media mix model with PyMC3 to optimize marketing spend allocation.
Aug 24, 2020 · Published elsewhere ↗
Beyond Recommendation Engines
Rethinking personalization beyond traditional recommendation systems at HelloFresh.
Jan 1, 2020 · Published elsewhere ↗
No matching articles. Try a different search.
No matching items