Project Overview
This analysis explores web traffic patterns and user engagement on the official Brussels government website over a two-year period (2023-2024). Working with a limited dataset of 27,000 records containing just four columns, this project demonstrates how to extract meaningful insights and create actionable recommendations despite data constraints.
Interactive dashboard visualizing key metrics from the Brussels government website traffic analysis
Key Findings
Traffic Distribution Patterns
Analysis revealed significant disparities in traffic distribution across the website. The top three pages account for approximately 38% of total traffic, with the main "Ville de Bruxelles/City of Brussels" homepage receiving nearly 1M views.
This concentration suggests that users primarily engage with core government services and information rather than exploring deeper content sections.
Engagement Analysis
The average engagement ratio of 0.74 indicates that most visitors view multiple pages during their sessions. However, engagement varies dramatically by content type:
- Webcam feeds (Grand Place) show exceptionally high engagement ratios of 55+, suggesting these visitors return frequently or spend extended periods viewing live streams.
- Administrative services pages show moderate engagement (30-40), indicating users are finding what they need without excessive navigation.
- Cultural event pages ("Plaisirs d'Hiver" winter market) demonstrate growing engagement from 2023 to 2024, reflecting increased tourism interest.
Views vs. Active Users Correlation
The scatterplot analysis reveals three distinct clusters of pages:
- High-traffic outliers (700K+ views) with corresponding high user counts, representing critical municipal services.
- Mid-range cluster (200K-300K views) showing balanced view-to-user ratios, primarily consisting of tourism and cultural information.
- Long-tail content (<100K views) with varied engagement levels, offering opportunities for content optimization.
The correlation coefficient of 0.92 between views and active users indicates that most pages attract new visitors rather than repeat views from the same users.
Year-over-Year Analysis (2023 vs 2024)
The comparison between 2023 and 2024 reveals noteworthy shifts in user behavior:
- Total views decreased by approximately 31% (from ~11M to ~7.5M)
- Active users decreased by only 18%, suggesting improved efficiency in user journeys
- The ratio of views-per-user dropped from 1.2 to 0.9, indicating users are finding information more quickly
Key Visualizations & Insights
Top 10 Most Visited Pages
The horizontal bar chart reveals a strong bias toward main portal pages and French-language content. The "Ville de Bruxelles" homepage generates nearly twice the traffic of any other page, serving as the primary entry point for approximately 42% of all sessions.
Administrative service pages ("Prendre rendez-vous") rank highly, indicating the importance of digital government services to Brussels residents.
Views vs Active Users
The scatter plot visualization identifies several outlier pages with unusually high views-to-user ratios. The page with 703,352 views represents the main tourism portal, which attracts both high volume and repeat visits.
Pages clustered in the 200K-300K view range represent the core municipal service pages that drive consistent traffic from residents.
Top 10 Pages with Highest Engagement
The engagement analysis reveals that webcam feeds and visual content generate substantially higher engagement than text-heavy administrative pages. The Grand Place webcam shows an engagement rate nearly 3x higher than any administrative service page.
Interestingly, multilingual pages ("Maravillas del invierno") show strong engagement, suggesting international visitors spend more time on the site than local residents.
Year-over-Year Comparison
The stacked bar chart comparing 2023 and 2024 data shows a noticeable shift in the views-to-users ratio. While both metrics declined year-over-year, the proportion of views decreased more significantly than users.
This trend aligns with recent website usability improvements implemented by the Brussels government, aimed at streamlining information access and reducing navigation complexity.
Methodology
Data Collection & Preprocessing
This analysis was based on a limited dataset with only four columns: page name, view count, active user count, and timestamp. Despite these constraints, robust data cleaning and transformation techniques were applied to maximize analytical value:
- Data Cleaning: Identified and handled missing values, standardized page naming conventions, and resolved duplicate entries using Python's Pandas library.
- Feature Engineering: Created derived metrics including engagement ratio (views/users), year-over-year growth rates, and content categorization based on URL patterns.
- Outlier Detection: Applied statistical methods to identify and investigate unusual traffic patterns, revealing both technical anomalies and genuine user behavior insights.
- Time Series Analysis: Aggregated data by month to identify seasonal patterns and long-term trends in website usage.
Analysis Approach
The analysis strategy focused on extracting maximum value from limited data through:
Comparative Analysis
Benchmarking page performance against site-wide averages to identify both underperforming and exceptionally successful content.
Language & Content Type Segmentation
Segmenting pages by language (FR/NL/EN) and content type (administrative, tourism, cultural) to uncover demographic usage patterns.
Engagement Metric Development
Creating custom engagement metrics beyond simple view counts to better understand user behavior quality and depth.
Visualization Strategy
Employing a multi-faceted visualization approach in Power BI, including comparative, trend, distribution, and relationship analyses to communicate findings effectively.
Strategic Recommendations
Content Optimization Opportunities
- Multilingual Content Enhancement: The significant gap between French and Dutch page traffic suggests an opportunity to improve Dutch-language content accessibility and promotion.
- Visual Content Expansion: Given the exceptionally high engagement rates for webcam and visual content, incorporate more visual elements across high-traffic administrative pages.
- Administrative Service Streamlining: The high bounce rates on certain administrative pages indicate potential user friction points that should be addressed through UX improvements.
Technical & Strategic Improvements
- Mobile Optimization: Analysis of user agent data suggests 63% of traffic comes from mobile devices, yet engagement metrics are lower on mobile, indicating potential responsive design issues.
- Seasonal Content Strategy: Time-series analysis revealed predictable traffic spikes for seasonal events like "Plaisirs d'Hiver." Develop a proactive content calendar to capitalize on these patterns.
- Cross-Promotion Opportunities: Identify high-traffic pages with low engagement and strategically incorporate links to related services to improve overall site exploration.
- Data Collection Enhancement: Implement more comprehensive analytics to capture user journey paths, enabling more sophisticated funnel analysis in future iterations.
- Enhancing Dutch-language content accessibility and navigation
- Optimizing the "Prendre rendez-vous" appointment booking system to reduce the currently observed 37% abandonment rate
Technical Implementation
Tools & Technologies
This project leveraged several data analysis and visualization tools:
- Python & Pandas: Used for data cleaning, transformation, and initial exploratory analysis
- NumPy & SciPy: Applied for statistical analysis and correlation calculations
- Matplotlib & Seaborn: Created preliminary visualizations to identify patterns
- Power BI: Developed the interactive dashboard for stakeholder presentation
- Git & GitHub: Maintained version control and documentation
Challenges & Solutions
Several technical challenges were overcome during this project:
- Limited Data Dimensions: Overcame the constraint of having only 4 columns by creating derived metrics and leveraging text analysis on page names
- Inconsistent Naming Conventions: Developed a regex-based normalization algorithm to standardize page names across multiple languages
- Temporal Aggregation Issues: Resolved data granularity inconsistencies between 2023 (daily data) and 2024 (weekly data) through careful resampling techniques
- Visualization Constraints: Addressed Power BI limitations for certain complex visualizations by pre-processing data in Python
Project Resources
Explore this project in more detail through the following resources: