How to Use EDA to Investigate the Effects of E-Commerce on Retail Real Estate

Exploratory Data Analysis (EDA) is a powerful approach to uncover patterns, detect anomalies, and test hypotheses using statistical graphics and other data visualization techniques. In the context of retail real estate, EDA can be instrumental in understanding the growing impact of e-commerce on traditional retail spaces. With the steady rise of online shopping, many brick-and-mortar stores are experiencing shifts in foot traffic, sales volume, and even closures—trends that directly influence retail property values, vacancy rates, and development strategies.

This article explores how to use EDA to investigate the effects of e-commerce on retail real estate using various data sources, visualization techniques, and statistical methods.

Identifying Relevant Data Sources

Before conducting EDA, identifying and compiling relevant datasets is critical. For analyzing the effects of e-commerce on retail real estate, consider the following sources:

• Retail Real Estate Data: Property values, lease rates, occupancy rates, square footage, tenant types.

• E-Commerce Sales Data: National and regional e-commerce sales trends by category.

• Consumer Behavior: Data on shopping preferences, foot traffic in shopping centers, and online search trends.

• Macroeconomic Indicators: Employment data, disposable income, inflation rates, and urbanization trends.

• Store Closure Announcements: Information on retail store closures or bankruptcies.

• Logistics and Fulfillment Center Growth: Data on warehouse leasing and last-mile delivery hubs.

Data can be obtained from government databases (e.g., U.S. Census Bureau, Bureau of Economic Analysis), private real estate firms (e.g., CBRE, JLL), and e-commerce industry reports (e.g., Statista, eMarketer).

Data Preprocessing for EDA

After collecting data, preprocessing ensures data quality and consistency. Common preprocessing steps include:

• Handling Missing Values: Fill, interpolate, or drop null values based on context.

• Data Type Conversion: Ensure correct data formats (e.g., datetime, categorical).

• Normalization: Scale data to allow fair comparisons, especially when comparing sales volume across store sizes.

• Feature Engineering: Create derived variables such as “E-commerce share of total retail,” “vacancy rate per region,” or “average lease duration by tenant type.”

Univariate Analysis: Understanding Distributions

Start EDA with univariate analysis to understand the distribution of individual variables:

• Histograms of retail property prices, lease rates, or square footage help assess distribution and skewness.

• Boxplots highlight outliers in vacancy rates across different cities.

• Bar charts can show frequency distributions of store closures by year or tenant category (e.g., apparel, electronics).

This helps identify trends like increases in closures or differences in lease prices by location or retail format (mall, strip center, standalone).

Bivariate Analysis: Detecting Relationships

Bivariate analysis reveals how two variables interact and is essential in understanding the effect of e-commerce on real estate metrics.

• Scatter plots can compare e-commerce penetration rates with retail vacancy rates. A positive correlation may suggest that higher online sales reduce demand for physical retail space.

• Line charts can track the growth of e-commerce sales alongside the decline in retail foot traffic or mall revenue over time.

• Heatmaps of correlation matrices help identify strong linear relationships among variables, such as between regional e-commerce growth and local retail property devaluation.

Analyzing these relationships can help determine if e-commerce is a primary factor in changing real estate conditions.

Time Series Analysis: Tracking Trends Over Time

Retail real estate and e-commerce are dynamic, evolving domains, so temporal analysis is crucial:

• Time series plots of national e-commerce sales vs. brick-and-mortar sales show divergence trends.

• Foot traffic analytics from sources like Placer.ai can demonstrate seasonality and long-term shifts in consumer visits to malls and shopping districts.

• Vacancy rate trends over time in urban vs. suburban areas can highlight structural changes in real estate demand.

Use rolling averages or decomposition to separate trend, seasonality, and noise in time series data.

Geographic Analysis: Spatial Insights

The impact of e-commerce is not uniform across regions. Geographic EDA can identify spatial patterns:

• Choropleth maps show retail vacancy rates or e-commerce sales penetration by state or metro area.

• Geospatial clustering identifies areas with rapid changes in property values or store closures.

• Zip-code level analysis can link local e-commerce adoption rates with retail leasing dynamics.

GIS tools like QGIS or Python libraries like geopandas and folium can be used to visualize and explore these spatial relationships.

Category-Level Comparison: Sector-Specific Impact

Different retail sectors face different levels of disruption from e-commerce. EDA can highlight these distinctions:

• Segmented bar charts compare lease rates and vacancy by retail category (e.g., grocery, fashion, electronics).

• Trend analysis shows how big-box stores vs. specialty retailers have fared over time.

• Cross-tabulations examine the proportion of vacant properties within each category, revealing which sectors are most vulnerable.

This helps developers, investors, and urban planners tailor strategies based on which sectors are more resilient.

Advanced EDA: Clustering and Dimensionality Reduction

For large datasets, use unsupervised learning to extract deeper patterns:

• K-Means Clustering: Group retail centers by characteristics such as size, tenant mix, and location to identify those most susceptible to e-commerce disruption.

• Principal Component Analysis (PCA): Reduce complexity and uncover latent variables affecting retail real estate performance.

Such techniques help summarize complex datasets and prioritize further analysis or predictive modeling.

Visual Storytelling: Dashboarding and Reporting

EDA is most effective when insights are communicated clearly. Use interactive dashboards to tell the story:

• Tools like Tableau, Power BI, or Python’s Dash and Plotly libraries help create dynamic visuals.

• Combine multiple charts to show interlinked trends—e.g., a dashboard that shows how foot traffic, e-commerce growth, and vacancy rates evolve together.

• Provide filters by time, geography, and retail type to allow stakeholders to customize their view.

Dashboards enhance accessibility for decision-makers such as investors, property managers, and urban planners.

Hypothesis Generation and Strategy Development

Ultimately, the goal of EDA is to move from observation to insight. Example hypotheses you can test:

• “Higher e-commerce penetration is associated with rising retail vacancies in suburban malls.”

• “Retail properties within 10 miles of Amazon fulfillment centers show lower appreciation.”

• “Foot traffic decline precedes lease renewal reductions by 6–12 months.”

These hypotheses can inform real estate investment decisions, zoning policy revisions, or adaptive reuse strategies (e.g., converting retail spaces into mixed-use developments or last-mile delivery hubs).

Conclusion

EDA is an indispensable tool in evaluating how the rise of e-commerce reshapes the retail real estate landscape. By leveraging statistical techniques, visualization tools, and robust datasets, analysts and stakeholders can detect patterns, understand causality, and make informed decisions. The insights gained through EDA not only clarify the extent of e-commerce’s impact but also highlight new opportunities for innovation in property development, urban planning, and retail strategy.