How to Detect Patterns in Online Shopping Cart Abandonment Using EDA

To detect patterns in online shopping cart abandonment, exploratory data analysis (EDA) is an effective approach. EDA helps in uncovering underlying structures, relationships, and insights that are critical in understanding the reasons behind cart abandonment. By identifying these patterns, businesses can make informed decisions to optimize their shopping experience and reduce cart abandonment rates.

Here’s how you can use EDA to detect patterns in online shopping cart abandonment:

1. Data Collection and Preparation

The first step in any EDA process is gathering the necessary data. For cart abandonment analysis, key variables might include:

• User information: Customer demographics (age, location, etc.).

• Cart details: Number of items, total cart value, category of items.

• Browsing behavior: Time spent on the website, number of pages viewed, clickstream data.

• Checkout progress: Whether the user proceeded through checkout steps like payment info or shipping details.

• Transaction details: Cart abandonment time, session duration, device used, and platform used.

Data preparation may involve:

• Removing or imputing missing data.

• Encoding categorical variables.

• Normalizing numerical data if required.

2. Univariate Analysis

Start with simple descriptive statistics to understand individual variables and detect any obvious patterns or anomalies in the data.

• Cart Value Distribution: Visualizing the distribution of cart values can help identify if higher-value carts are more likely to be abandoned.

• Time on Site: Plotting histograms of the time spent on the site helps uncover if users who abandon carts tend to spend more or less time on the website.

• Abandonment Rate: Calculate the overall abandonment rate and observe how it varies with different categories (e.g., product categories, payment methods, etc.).

Box plots, histograms, or bar charts are useful for summarizing these variables.

3. Bivariate Analysis

Once individual variables are understood, the next step is to examine relationships between variables. This can highlight potential correlations and dependencies that lead to abandonment.

• Cart Value vs. Abandonment: Does the cart value correlate with abandonment? A higher cart value could indicate that users abandon the cart more often due to concerns over pricing.

• Time on Site vs. Abandonment: Does a longer time spent on the website correlate with abandonment? It might suggest users are overwhelmed by too many choices or are distracted by unrelated elements.

• Number of Products vs. Abandonment: Users with larger carts may abandon more frequently due to the complexity of managing larger orders.

• Device/Platform: Are certain devices (mobile vs. desktop) or platforms (iOS vs. Android) associated with higher abandonment rates?

Use scatter plots, heatmaps, or pair plots to analyze these relationships visually.

4. Categorical Analysis

• Payment Methods: Analyze if certain payment methods (credit card, PayPal, etc.) have a higher abandonment rate.

• Shipping Options: Are certain shipping options correlated with abandonment? For instance, higher shipping fees might discourage finalizing a purchase.

• Geographic Trends: Check if abandonment rates vary by user location. For example, international users might abandon carts more frequently due to shipping or currency concerns.

Bar charts and stacked bar plots can be helpful here.

5. Time Series Analysis

Cart abandonment may have temporal patterns, such as:

• Day of the Week: Are users more likely to abandon carts on weekdays or weekends?

• Time of Day: Does abandonment peak during certain hours, suggesting time-sensitive issues like server performance, checkout time, or user behavior based on working hours?

• Seasonality Trends: Cart abandonment might increase during certain months (e.g., after holidays) or even certain seasons (e.g., after sales events).

Time series plots or line charts can be used to detect any time-related patterns.

6. Segmentation Analysis

Segment the data into different user groups to better understand which factors might be influencing abandonment within different user segments:

• New vs. Returning Users: Are new users more likely to abandon their carts compared to returning users? New users might not trust the site or feel less committed to purchasing.

• User Behavior: Segment users based on their browsing behavior (e.g., those who browsed multiple product categories vs. those who focused on a specific one). This can help you understand if users with specific interests are more likely to abandon their carts.

Use group-by operations in pandas or SQL to create meaningful segments and visualize the differences.

7. Detecting Outliers and Anomalies

Identify extreme cases that might skew the data, such as users who add extremely high-value items to the cart or spend an unusually long time on the site. These outliers may highlight issues that cause cart abandonment (e.g., technical glitches or unusual user behavior).

Outlier detection methods, such as IQR (Interquartile Range) or Z-scores, are helpful to identify these extreme cases. Box plots are commonly used for visualizing outliers.

8. Correlation Analysis

Correlation analysis can provide insight into relationships between numerical variables, such as cart value, session time, number of items, and abandonment.

• Pearson Correlation Coefficient: This measures the linear relationship between two variables. For example, a negative correlation between session time and abandonment could suggest that users who spend less time on the site are more likely to abandon their carts.

• Spearman Rank Correlation: This non-parametric test is useful for understanding monotonic relationships, especially when the data isn’t normally distributed.

A heatmap of the correlation matrix can help visualize these relationships easily.

9. Hypothesis Testing

Use hypothesis tests to statistically validate the findings from the EDA. For example:

• Is the average cart value higher for abandoned carts compared to completed transactions?

• Do users who interact with certain product categories have a significantly higher abandonment rate?

Conduct tests like t-tests or ANOVA to validate these assumptions.

10. Machine Learning Insights (Optional)

After completing EDA, machine learning models can be used to predict cart abandonment. Common approaches include:

• Logistic Regression: A simple model to predict abandonment (binary classification).

• Decision Trees: To uncover non-linear patterns in the data.

• Clustering: Unsupervised learning to find groups of similar users who are more prone to abandon their carts.

While this goes beyond pure EDA, it can enhance the understanding of patterns when combined with insights from EDA.

Conclusion

By applying EDA to cart abandonment data, businesses can unearth valuable patterns and gain insights into user behavior. These findings can help optimize the shopping experience, reduce abandonment rates, and ultimately increase conversions. EDA reveals the “why” behind abandonment, whether it’s due to pricing, shipping issues, user experience problems, or other factors. By addressing the root causes, businesses can improve their retention and sales figures.