🌐 Live Dashboard: https://ukpropertydashboard-gtgbfxutdfmphrvmvhmtbw.streamlit.app
Property prices are influenced by multiple factors, yet buyers, sellers, and investors often rely on asking prices or intuition when making decisions.
To address this, I developed an interactive analytics dashboard that analyzes 7,999 UK residential property transactions between 2010 and 2025 and uses machine learning to estimate property prices based on historical market patterns.
The solution combines business intelligence, exploratory data analysis, and predictive analytics to help users understand what truly drives property values rather than relying on assumptions.
Business Problem: Residential property pricing is rarely determined by a single factor.
Questions such as:
- Which locations consistently command premium prices?
- Does property type significantly affect value?
- How much does Freehold vs Leasehold ownership influence pricing?
- Are price changes caused by increased demand or simply different types of properties being sold?
- Can historical transaction data be used to estimate property value?
are difficult to answer without structured analysis.
This dashboard provides a data-driven approach to understanding market trends by identifying the characteristics that consistently influence pricing while offering predictive estimates for future decision-making.
The dashboard was designed to:
- Identify the key drivers of residential property prices.
- Compare pricing trends across districts, property types, and tenure.
- Analyse historical market behaviour between 2010 and 2025.
- Predict property prices using machine learning.
- Enable users to interactively explore the housing market through dynamic visualisations.
- Python
- Pandas
- NumPy
- Scikit-learn
- Plotly
- Streamlit
This project showcases practical experience in:
- Data Cleaning
- Exploratory Data Analysis (EDA)
- Feature Engineering
- Data Visualisation
- Machine Learning
- Predictive Analytics
- Business Intelligence
- Dashboard Development
- Statistical Thinking
- Business Communication
- Stakeholder Reporting
Source: Client-provided proprietary UK property transaction data.
Records: 7,999 property transactions
Time Period: 2010–2025
File Format: CSV
- Sale Price
- Date of Transfer
- Property Type
- Tenure
- New Build Status
- District
- Town/City
- County
Before beginning the analysis, I performed a full data quality assessment to understand potential issues that could affect insight accuracy.
Missing District value: 984 Duplicate records:35 High-value outliers (> £671,500): 529 Maximum recorded transaction: £26.7M
Additional observations:
- No negative or zero property prices
- Categorical variables contained consistent values
- Date fields were standardized for time-series analysis
The dataset was prepared using Python and Pandas.
The workflow included:
- Removing records outside the project scope
- Handling missing and duplicate records
- Converting transaction dates into Year and Month features
- Detecting high-value outliers for analytical interpretation
- Encoding categorical variables for machine learning
- Preparing features for regression modeling
The analysis focused on identifying the variables with the greatest influence on property prices.
The dashboard enables users to explore:
- Average price distribution by district
- Average prices by property type; Freehold vs Leasehold comparison
- Yearly transaction trends
- Sales volume over time
- Geographic pricing differences
_Interactive filters allow users to compare districts, property categories, and historical trends in real time.
-
Location is the strongest pricing driver Districts such as Crawley, Camden, Westminster, and Hammersmith & Fulham recorded the highest average property prices. However, several premium districts had relatively few transactions, meaning exceptionally high averages should always be interpreted alongside transaction volume.
-
Property type significantly impacts value Detached properties achieved substantially higher average selling prices than other property categories.
Detached: ~£412K
Flat : ~£257K
Terraced: ~£255K
This demonstrates that reporting one overall average property price masks meaningful differences across the market.
-
Freehold properties consistently command higher prices Freehold properties consistently sold for more than Leasehold properties.
Average Freehold price: ~£320K
Average Leasehold price: ~£250K
This represents an approximate 25% pricing premium, suggesting tenure is an important factor in property valuation.
- Transaction volume remained relatively stable Average property prices experienced noticeable fluctuations over time, including a decline around 2020 before recovering by 2025.
However, annual transaction volume remained relatively consistent at approximately 450–500 transactions per year.
This suggests that changes in average prices were driven more by the mix of properties being sold than by overall market activity.
To complement the descriptive analysis, I built a regression model capable of estimating residential property prices.
Model: Regression (Scikit-learn)
Features:
- District
- Property Type
- Tenure
- Transaction Year
Property Price
The trained model is integrated into the Streamlit application, allowing users to generate estimated property prices based on selected property characteristics.
The dashboard includes:
- Interactive filtering
- District comparison
- Historical price trends
- Transaction volume analysis
- Property type comparison
- Tenure analysis
- Property price prediction using machine learning
Based on the findings, the following recommendations were made:
- Segment market analysis by property type rather than relying on overall averages.
- Make tenure a primary decision variable during property valuation.
- Pair price metrics with transaction counts to avoid misleading conclusions from low-volume districts.
- Flag districts with limited sales activity so unusually high averages are interpreted appropriately.
- Use predictive pricing as a supporting decision tool rather than a replacement for professional valuation.
This solution demonstrates how analytics can transform historical transaction data into practical decision-support tools.
- Potential applications include:
- Helping buyers compare asking prices with historical market trends.
- Supporting sellers in setting more competitive listing prices.
- Assisting investors in identifying high-value locations.
- Providing estate agencies with data-backed pricing insights.
- Offering stakeholders an interactive alternative to static market reports.
This analysis is constrained by the available dataset.
- The model does not account for:
- Property condition
- Floor area
- Number of bedrooms or bathrooms
- Property age, Interest rates
- Inflation
- Mortgage rates
- Other macroeconomic factors
Additionally, district averages may be influenced by low transaction volumes.
This project demonstrates an end-to-end analytics workflow, from raw data preparation through business insight generation and predictive modelling.Rather than simply visualising historical transactions, the dashboard explains the factors that influence residential property prices and provides an interactive prediction tool that supports evidence-based decision-making.
