The Marketing Science Signal
Applied marketing science and AI, from model to revenue
Recent Posts
- Human-in-the-Loop: Data Quality for AI
- The Cold Start Problem: Why Prospects Need Their Own Propensity Model
- Generative Engine Optimization (GEO) for Marketing Impact
- From Score to Action: Operationalizing Pipeline Intelligence — Pipeline Optimization Part 3
- Data Is Noisy: How Machine Learning Solves the “Garbage In, Garbage Out” Problem
About the Author
Mike is a leader in the field of Marketing Data Science & Operational Strategy with 20+ years leading global Data Science, AI/ML, and Marketing Analytics teams at Dell Technologies, Cisco, Pure Storage, Hitachi Vantara and Hearst Media. He is also an Accredited Professional StatisticianTM with the American Statistical Association.
Michael Foley, PStat®
Marketing data science, AI and analytics leadership that turns models into revenue.
20+ years leading global data science, AI/ML and marketing analytics teams at Dell Technologies, Cisco, Pure Storage and Hitachi Vantara. Accredited Professional Statistician (American Statistical Association), M.S. Predictive Analytics (Northwestern), MBA (Haas). These articles are my working notes on propensity, segmentation, forecasting, pipeline optimization and AI data quality.
-
Why data quality still decides whether AI can be trusted: EDA, data dictionaries, and separating “no data” from “pass” in automated checks.
-
A blended model hides the prospect gap. Splitting customers from prospects reveals a gap nine times larger, and why prospects need their own model.
-
How generative engine optimization changes content discovery, and what marketers can do so AI assistants find and cite their work.
-
Part 3: turning purchase-likelihood and segmentation models into a production revenue engine, with leakage audits, GTM thresholds, and forecast impact.
-
Why waiting for perfect data costs more than noisy data, and how machine learning handles “garbage in, garbage out.” From my conversation with Nav Thethi.
-
Part 2: building a propensity model that predicts Closed Won from B2B Salesforce pipeline data, and using the score to prioritize deals.
-
Part 1: a statistical framework for B2B pipeline health, using funnel conversion, deal velocity, and RFM to see up to three quarters ahead.
-
How look-alike behavior solves the discovery problem in B2B and B2C cross-sell. Part 3 of my recommender systems series.
-
Building an AI agent that extends my marketing methodology, moving from competing with agentic AI to integrating it as scalable staff.
-
Modeling the customer journey with Markov chains to predict the next likely purchase from the current state. Part 2 of my recommender systems series.
-
Association rules for cross-sell, moving high-potential customers toward Champions. Part 1 of my recommender systems series.
-
I pitted a Gemini-based AI agent against my own reading of an XGBoost propensity model. AI won on scale; human judgment won on context.
-
Combining field sales expert opinion with statistical and deep learning ensembles for more accurate B2B revenue forecasts.
-
How coordinating digital nurture with direct outreach triples conversion lift, going beyond the old “3+ exposures” rule.
-
A framework for B2B contact targeting: when contact-level data matters, and how to build personas beneath the account model.
-
Using XGBoost and survival analysis on the IBM Telco dataset to identify at-risk customers and predict when they will churn.
-
Moving beyond RFM to predictive customer lifetime value, using segmentation and propensity scores as inputs for account prioritization.
-
Why propensity-to-buy models are the workhorse of demand generation, and how binary classification focuses targeting.
-
Why RFM remains my favorite segmentation scheme for B2B account portfolio management, from BCG’s matrix to machine learning.