Oberon Analytics

What I do

Advisory

Fractional analytics leadership

Interim or fractional head of analytics and insights. Set the agenda with your leadership team, fix the foundations (definitions, governance, one source of truth), and stand up a team or a practice you keep after I leave.

Measurement

Experimentation and causal measurement

Design and run the tests that settle arguments: pricing and fee experiments, feature tests, lifecycle and retention levers, incrementality. Doctoral training in experimental design, twenty years of applying it to customers, consumers, and workforces.

AI enablement

AI-assisted analytics that a risk officer would sign

Move an analytics team from spreadsheets and point tools to Python and LLM-assisted workflows, under a risk-tiered framework for sensitive data. Governance and productivity at the same time, because that is the only version that survives an audit.

How I work

Decision first. Every engagement starts with the decision you need to make and the number that would change it.
Numbers with denominators. Baselines, comparison groups, and honest uncertainty. No "doubled" without a denominator.
Built to be kept. Definitions, code, and process are handed over, documented, and yours.
Hands on. I write the SQL and the Python. I also present to your board.

Proof

Built an enterprise analytics function. Inherited a request-driven reporting team inside a $6B business and rebuilt it into a research and data science function at flat headcount; the parent enterprise adopted the model across 26 business units and ~$23B in revenue.
Put a price on a decision. Designed transaction-fee experiments measuring funnel loss at each price point and built the revenue model behind a ticketing platform's decision to introduce fees, projecting annual revenue from $30M to $60M+ at flat volume.
Made the forward view real. Attrition-forecasting models for a 50,000-person workforce whose top risk tier ran at 1.65x the baseline rate on live post-deployment data; leaders acted on it.
Moved a team to modern tooling in weeks. Converted a 10-analyst team from Alteryx to Python in two months after a year of formal training had converted two, retiring $70K a year in licensing; cut a 7,000-comment survey analysis from a week to a day with an LLM pipeline.
Shipped a product, solo. stathead.app: an open NFL analytics platform (~125K lines of TypeScript and Python, 30-tool MCP server on npm with 2,500+ monthly downloads, weekly model retraining) built in six months with agentic AI.

Insights

Coming

A risk-tiered framework for AI use on sensitive data

How an analytics team can adopt LLM tooling on HR data without a compliance incident, mapped to the NIST AI Risk Management Framework.

Coming

Marketing mix modeling on open data

A worked Meridian model with calibration against a holdout, and what the results can and cannot tell a CMO.

Coming

When your forecasts quietly expire

A production incident from stathead.app and the monitoring that fixed it: a case study in the Measure and Manage functions.

About

Matt Porritt, PhD
Founder, Oberon Analytics
Atlanta, GA

Twenty years leading analytics and data science: most recently Senior Director of Data Analytics at Cox Enterprises, before that building analytics from zero for a live-event ticketing start-up, and a decade of customer and consumer modeling for a 6M-customer cable business, automotive media properties, and agency clients in food service, retail, telecom, and automotive.

PhD in applied behavior analysis from Western Michigan University, with concentrations in experimental design and organizational performance. Peer-reviewed publication; invited keynote on consumer behavior.

Contact

Engagements are scoped, fixed-fee where possible, and start with a 30-minute conversation about the decision you are trying to make.

mlporritt@gmail.com LinkedIn