Open to junior Data Analyst and Marketing Analyst roles anywhere in the UK, or remote. Right to work: UK Graduate visa, December 2026 to December 2028. No employer sponsorship required.
Each project lists its limits. Simulated data is labelled as simulated.
| Project | What I found | Stack |
|---|---|---|
| H&M customer retention | Ranking at-risk customers by risk x past spend caught 49.4% of lost revenue. A plain churn model caught 11.0% from the same list size. | Python, DuckDB, Tableau, Ollama |
| F1 2026 championship forecast | Two models, two answers: Antonelli at 22.4% in the weighted SQL model, 76.1% in 5,000 Monte Carlo runs. The gap is the interesting part. | BigQuery SQL, Python |
| Tesco vs Ocado online grocery | Ocado averaged 33.8 on Lighthouse performance against 49.8 for Sainsbury's. 10 experiments scored on ICE. | Meta Ad Library, Lighthouse, Tableau |
| Revolut London growth study | Simulated data. 59 of 500 campaigns had negative ROI, about £1.9M of £12.7M simulated spend. | Python, SQL |
| Harry Styles tour analytics | Revenue concentrates in New York and Los Angeles. Stadiums vary more in revenue per seat than arenas. | Python, SQLite, Plotly |
| BNPL and impulse buying | MSc dissertation, 163 UK shoppers. BNPL use predicted impulse buying (beta 0.472, p < 0.001). Correlational, not causal. | SPSS, regression |
| Languages | Python (pandas, NumPy, scikit-learn), SQL (window functions, CTEs, DuckDB, SQLite, BigQuery) |
| Analytics | Excel, Tableau, GA4, A/B test design, regression, SPSS |
| Marketing | Meta Ads Manager, Google Ads (Search certified), HubSpot |
| AI | Local LLMs with Ollama, plus number checkers for AI-written reports |
- Start with the business question, not the dataset.
- Check the sample is fair before trusting the result.
- Keep the model simple when a complex one scores the same.
- Test the AI. My first AI-written summary mixed up customers and revenue, so I built a number checker.
- Write the limits down. Accuracy is not the same as usefulness.
I started in filmmaking and marketing, and the question that kept coming back was which of it actually worked? I finished an MSc Marketing (Merit) at the University of Birmingham while teaching myself SQL, Python and machine learning on the side. That mix of analytics and marketing is where I want to work.