Michael Belli

Analytics Leader · 20+ Years Experience · Data-Driven Strategy

LinkedIn GitHub Email

Welcome

I’m an analytics leader who turns marketing and customer data into decisions executives can act on. This portfolio is a small set of self-directed, end-to-end projects — each built on synthetic data I generated, so the methodology is fully visible and reproducible from raw data through to a business recommendation.

Each project is meant to show how I think, not just what I can model: how I frame the business question, where the analysis can mislead, and what the honest takeaway is. Pick a project below to dig in.

A note on the data: Every figure here is an output of a simulation I wrote, not a client engagement. Working from a known ground truth is deliberate — it’s the only way to check whether a model actually recovers the truth, which is the whole point of the second project.

Projects

Insurance Marketing Analytics Decision Engine

An end-to-end analytics workflow for a B2C insurer — exploratory analysis, a Hill-saturation marketing mix model, profit-maximizing budget optimization, and business-impact measurement. The throughline: in insurance, cheap leads often become unprofitable policies, so growth has to be measured in profit and lifetime value, not lead volume.

Explore the project →

When a Marketing Mix Model Fits Well and Still Points the Wrong Way

A focused cautionary tale: a marketing mix model can post a strong fit — high R², low error — and still recommend cutting your better channel. Here the culprit isn’t the model; it’s a billing-date timing lag in the data. Built on a known ground truth, the project shows four increasingly sophisticated models all getting the budget call backwards, and why the fix lives in the data, not the model.

Explore the project →
Click to open Click to open

Direct Mail Targeting: Mail Less, Make More

A simulated 100,000-household direct-mail campaign showing what a targeting model is worth in dollars. Scoring the file into deciles and mailing only where incremental sales cover the postage cuts pieces 40% and more than doubles ROI (45% → 102%) — the model isn’t just a prediction, it’s a stopping rule that prices out exactly where in the file to stop mailing. Explore the project →
Click to open

About Me

I’m an analytics leader with 20+ years of experience developing data-driven strategies to optimize business performance. My work spans predictive modeling, marketing analytics, customer segmentation, and executive decision support.

Core Skills:

Let’s Connect

I’m open to analytics leadership roles and consulting engagements.