← Back to projects Shipped

UK Property Deal Analyser

Which UK areas actually stack up for a buy-to-let investor right now? This project answers that with public data instead of opinion: 2.9 million HM Land Registry transactions cleaned and aggregated in SQL, joined to ONS rent statistics, scored with a transparent weighted model, and served as a Power BI deal-finder dashboard.

View code on GitHub

Why this project

Property deal sourcing runs on gut feel and hearsay. The data to do better is free: every sale in England and Wales is published by HM Land Registry, and the ONS publishes average rents by local authority. This is the dashboard a deal sourcer or first-time investor would actually pay for, built end to end from raw public data, and every number in it is reproducible from two commands.

What the data says

The North beats the South on every investor metric. Liverpool tops the deal score: 18.1% price growth since 2023, a £183k median entry price and a 5.6% gross yield. Newcastle upon Tyne has the highest gross yield of any major district at 6.5%.
Small northern flats are the yield play. Sunderland flats return 9.9% gross; Middlesbrough flats 8.8%. Kensington and Chelsea returns 4.0% on a £1.08m median price.
Prime London is repricing. Westminster median prices fell 13.7% and Hammersmith and Fulham 15.7% over the window, while Burnley and Liverpool led growth.
New builds carry a median 28% premium over established stock in the same district.

How it works

SQL pipeline2.9M raw rows loaded into DuckDB; cleaning and aggregation in plain, reviewable .sql files (2.4M clean sales out)

↓

Rent joinONS average rents by district and property type; 314 of 316 districts matched via name normalisation

↓

Deal scoreWeighted z-scores: yield 40%, growth 25%, affordability 20%, liquidity 15%; weights tested for ranking stability

↓

OutputsCharts, a ranked deal table, and a star schema of CSVs feeding the Power BI dashboard (model, DAX measures and page spec in the repo)

Honest limitations

The score uses gross yield, not net; rents are advertised-stock averages; and the ranking is a shortlist generator, not financial advice. Districts with under 300 sales a year are excluded because a thin market makes both the median and the exit unreliable. All of this is documented in the repo rather than hidden.

Skills demonstrated

SQL at scale: window functions, quantiles, dedup logic over millions of rows in DuckDB
Data cleaning with documented, defensible decisions
Power BI modelling: star schema, DAX measures, drill-through dashboard design
Business storytelling: findings a non-technical investor can act on
Reproducibility: unit-tested pipeline, CI, every figure regenerated from source