🚀 Introducing the Ultimate A/B Testing and Experiment Analysis Dataset! 👀
Are you ready to dive into some high-impact A/B testing and conversion optimization? 📊 If you're working on data science, marketing analytics, or just love exploring real-world business concepts, this dataset is perfect for you!
🔍 What’s Inside?
- 30,000+ sessions with detailed user behavior: time spent, pages visited, and more!
- Demographic breakdown: Age, gender, location, and device usage.
- Conversion data: Sign-ups, purchases, and revenue generation from different landing page designs (Vibrant, Cold, Heat).
- Engagement insights: Understand how factors like device type, session duration, and traffic source affect conversions.
📈 Why Should You Use This Dataset?
- Perfect for performing A/B testing, analyzing conversion rates, and exploring user segmentation.
- Explore multivariate analysis, hypothesis testing, and statistical significance for simulated business insights.
- Practical application: Ideal for learning, experimentation, or creating models that improve sales and marketing performance.
🌟It mirrors real-world scenarios to help you practice and enhance your skills. Whether you’re a beginner or an expert, this dataset will help you understand conversion analysis, optimization, and marketing strategies with data you can trust for learning. 💥
🔗 Ready to get started? Check out the dataset now and join the discussion! 💬 Let’s explore the secrets behind successful conversions together. 🚀
https://www.kaggle.com/datasets/sandeep1080/bassburst/data
| Column Name | Description |
|---|---|
user_id |
Unique identifier for each visitor. |
session_id |
Unique identifier for each session or visit. |
sign_in |
Indicates if the user logged in via email or used guest access. |
name |
Name of the visitor (generated from multiple locales). |
demographic_age |
Age of the visitor (ranging from 14 to 80). |
demographic_age_group |
Age group of the visitor: "Teenage", "Adult", "Old". |
demographic_gender |
Gender of the visitor (Male, Female, Not Answered). |
email |
Email address of the user (if logged in via email). |
location |
The city where the visitor is located. |
country |
Country corresponding to the location. |
device_type |
Type of device used (Mobile, Desktop, Tablet). |
timestamp |
Session start time. |
variant_group |
The landing page design variant the user saw (Vibrant, Cold, Heat). |
time_spent |
Total time (in minutes) the user spent on the landing page. |
pages_visited |
Number of pages the user viewed during the session. |
conversion_flag |
Binary flag (0/1) indicating whether the user converted (signed up or made a purchase). |
conversion_type |
Type of conversion achieved (Signup or Purchase). |
traffic_source |
Source of traffic (Organic, Paid, Social, Referral). |
product_purchased |
List of products purchased (if applicable). |
revenue |
Total revenue generated from the transaction (if purchase was made). |
payment_type |
Payment method used (Card or COD). |
card_type |
Card type if payment was made by card (Amex, Visa, Master). |
coupon_applied |
Whether a coupon was applied (Yes/No). |
bounce_flag |
Indicates if the session was a bounce (only one page visited). |
| Algorithm/Model | Description |
|---|---|
| A/B Testing | Used for comparing landing page variants and determining the most effective design. |
| Logistic Regression | For analyzing the relationship between various factors (e.g., time spent, pages visited) and conversion outcomes. |
| T-Test | Used to compare the means of time spent and pages visited between converters and non-converters. |
| ANOVA | Used to compare pages visited across different demographic age groups. |
| Chi-Square Test | Used to evaluate the association between categorical variables such as sign-in type and conversion outcomes. |
| Kaplan-Meier Estimator | For survival analysis, to estimate time-to-conversion. |
| Sequential Testing | Monitoring conversion rates over time with cumulative Z-tests. |
| Bayesian Methods | Posterior distribution analysis for conversion rates. |
| Multi-Armed Bandits | Thompson Sampling for dynamically selecting the best-performing landing page variant. |
| Library | Purpose |
|---|---|
| Pandas | For data manipulation and handling. |
| NumPy | For numerical operations and array manipulation. |
| Matplotlib/Seaborn | For data visualization (plots and graphs). |
| Statsmodels | For statistical models and hypothesis testing. |
| Scipy | For statistical tests like T-tests and Chi-Square. |
| Lifelines | For survival analysis (Kaplan-Meier). |
| Plotly | For interactive visualizations. |
| Statsmodels.formula.api | For fitting logistic regression models using formulas. |
| Statsmodels.stats.proportion | For calculating proportions and effect sizes. |
| Warnings | For suppressing unnecessary warnings during analysis. |
Throughout the analysis, we addressed 23 key business questions that helped us uncover actionable insights for optimizing the landing pages and user experience. These questions were centered around understanding:
- User Engagement: How are users interacting with the website?
- Conversion Metrics: Which landing page variant and user behaviors are driving conversions?
- Demographics Impact: How do age, gender, location, and device type influence conversion rates?
- Marketing Strategy: Which traffic sources (organic, paid, referral) perform best?
- Revenue Analysis: How does session behavior impact revenue generation and average order value?
- User Segmentation: What are the behaviors of first-time visitors vs. returning users?
- Time & Session Metrics: What is the impact of time spent and pages visited on conversions?
Each of these business questions was analyzed using various statistical methods and techniques:
- Exploratory Data Analysis (EDA): To understand the distribution of data, conversion rates, and behavior across different segments (age, gender, device type).
- A/B Testing: To compare landing page variants and determine the most effective one in terms of conversion rates.
- Hypothesis Testing: Employed tests like two-proportion z-tests, T-tests, ANOVA, and chi-square tests to assess statistical significance.
- Regression Analysis: Used logistic regression to analyze the relationship between user behaviors (time spent, pages visited) and conversion rates.
- Multivariate Testing: To account for multiple factors affecting conversions, including demographic and engagement variables.
- Bayesian Methods & Multi-Armed Bandits: To dynamically evaluate which landing page variant performs best over time.
- Sequential Testing: Analyzed conversion rates as sample sizes increased to ensure stable and reliable results.
To see the full analysis, experiment with different questions, and explore how each metric impacts business performance, you can find the code in the following links:
- Kaggle Dataset: Kaggle Link
- Kaggle Code: Kaggle Link
🤝Thank you!