Global IQ: Outcomes
An industry-first variable Bayesian MMM for a media owner. One PyMC framework across a growing brand portfolio, with adstock, saturation, hierarchical priors and automated quality checks.
PyMC · MLflow · Snowflake · AWS
Marketing science · Data science · London
I build Bayesian marketing-mix models, and the systems that put them to work.
Marketing scientist and data scientist with a first-class degree in Mathematics with Economics. I designed and launched Global IQ: Outcomes, an industry-first variable MMM for a media owner, and built the tracking, dashboards and AI tools around it.
I specialise in Bayesian Marketing Mix Modelling, and I own the whole path from a probabilistic model to a system running in production.
At Global Media Group Services I designed and launched Global IQ: Outcomes, an industry-first variable MMM for a media owner. It is a single Bayesian framework in PyMC that serves 45 brands live across Audio and Out-of-Home advertising, and is engineered to scale toward around 500.
I built it from first principles rather than an off-the-shelf framework, after reviewing Google Meridian, Meta Robyn and PyMC-Marketing and finding none could handle the variable, multi-brand requirement. Around the models I built the working parts too: experiment tracking, automated quality checks, interactive dashboards, and a self-serve assistant that lets colleagues ask the results questions in plain language.
I enjoy the point where careful statistics meets clean engineering, and where a model turns into a decision someone can actually use.
Skills
Measuring what marketing genuinely drives, and where extra spend stops paying off. Built in PyMC with adstock and saturation curves, hierarchical priors and ROI estimation. Familiar with Google Meridian, Meta Robyn and PyMC-Marketing.
Telling real effect from coincidence, and grounding models in evidence. Lift-test calibration, incrementality, geo-experiment design, attribution and unified measurement.
Probabilistic models that carry their own uncertainty. Bayesian inference and sampling (MCMC), time-series methods, regularised regression, and rigorous model checking.
Getting models off the laptop and running reliably at scale. Experiment tracking and a model registry with MLflow, automated quality checks, parallel training, and CI/CD.
Making results self-serve through AI assistants. Building agents and tools with the Model Context Protocol, retrieval over internal data, and reusable prompts and skills in Claude Code.
Clean data, in the cloud, ready to model. Python and SQL, Snowflake and dbt pipelines, and AWS (S3, Lambda, DynamoDB, CloudFront).
Experience
Career break, Mar to Aug 2025: travel through Japan and Southeast Asia, with self-directed study in AWS and Bayesian methods.
Selected work
An industry-first variable Bayesian MMM for a media owner. One PyMC framework across a growing brand portfolio, with adstock, saturation, hierarchical priors and automated quality checks.
An AI assistant that pulls live Formula 1 data through the Jolpica API to reason about race outcomes. A hands-on study in agent tooling and the Model Context Protocol.
This site, hand-coded in HTML, CSS and JavaScript and served on S3 and CloudFront, plus serverless experiments with Lambda and DynamoDB behind a small API.
Education
Certifications & interests
Music. Grade 8 saxophonist with the ABRSM, fifteen years of practice.
Chess. Long-time competitive player, drawn to the strategy.
Contact
Open to marketing science, MMM and data science roles, in consultancy or in-house. The quickest way to reach me is email.