End-to-end analytics project based on the Olist Brazilian E-Commerce dataset. The project analyzes marketplace performance, revenue drivers, delivery operations, customer satisfaction, customer retention, and segmentation. The final output includes a Jupyter Notebook analysis, BigQuery SQL queries, and interactive Tableau dashboards.
The goal of this project is to analyze Brazilian e-commerce marketplace performance and identify key factors affecting: revenue growth, product category performance, delivery efficiency, customer satisfaction, customer retention, and long-term business opportunities.
- Python: data cleaning, feature engineering, EDA, statistical testing, RFM segmentation, cohort analysis
- Pandas, Matplotlib, Seaborn, SciPy: analysis and visualization
- BigQuery / SQL: business question validation and analytical queries
- Tableau Public: interactive dashboard development
- Source: Kaggle — Olist Brazilian E-Commerce
- 99,441 orders, 8 tables
- Period: September 2016 — August 2018
- Validated analysis period: January 2017 — August 2018
- Scope: orders, customers, sellers, products, payments, reviews, and geographic data
Incomplete historical months were excluded from the main analysis:
2016-10 — partial first month in the dataset
2016-12 — only one recorded order
- Data loading and initial exploration
- Data cleaning and validation
- Feature engineering
- Master table creation
- Business performance analysis
- BigQuery SQL validation
- Customer segmentation and cohort analysis
- Tableau dashboard development
- Business conclusions and recommendations
brazilian-ecommerce-analysis/
│
├── data/
│ ├── README.md
│ ├── clean_df.zip
│ ├── cohort_retention_tableau.csv
│ ├── repeat_customers.csv
│ └── rfm.csv
│
├── notebooks/
│ ├── brazilian_e-commerce_project.ipynb # Full consolidated notebook (all stages combined)
│ ├── 01_data_preparation.ipynb # Data loading, cleaning, feature engineering, master table
│ ├── 02_business_analysis.ipynb # Business performance analysis
│ └── 03_retention_segmentation.ipynb # Customer segmentation and cohort analysis
│
├── queries.sql
│
├── visualizations/
│
└── README.md
The analysis is available as a single consolidated notebook (brazilian_e-commerce_project.ipynb) and split into three separate files for GitHub preview compatibility.
- How did revenue and order volume change over time?
- Which product categories generated the most revenue?
- Which states contributed the most to revenue and customer activity?
- Which sellers generated the highest revenue and customer satisfaction?
- Which payment methods were most popular?
- How did delivery delays affect customer review scores?
- Were late deliveries statistically associated with lower ratings?
- Which states had the longest delivery times?
- Which product categories received the lowest review scores?
- What share of customers made repeat purchases?
- Which RFM customer segments were the largest?
- How did cohort retention change over time?
BigQuery was used to answer and validate core business questions using SQL.
The SQL file includes analytical queries for:
- revenue by product category,
- monthly revenue and order volume,
- average delivery time by state,
- revenue contribution by state,
- payment method analysis,
- lowest-rated product categories.
These queries are aligned with the main business sections in the Jupyter Notebook and support the Tableau dashboard datasets.
- The platform showed strong growth throughout 2017.
- Revenue peaked in November 2017, likely driven by Black Friday demand.
- By 2018, monthly revenue stabilized at a high level.
- The top product categories included health_beauty, watches_gifts, and bed_bath_table.
- Revenue was geographically concentrated, with São Paulo generating the largest share.
- Late deliveries were strongly associated with lower customer review scores.
- On-time deliveries had an average review score of 4.21.
- Late deliveries had an average review score of 2.26.
- A statistical test confirmed that the difference was significant.
- Northern and remote states had longer average delivery times.
- Repeat purchase rate was very low at approximately 3%.
- Around 97% of customers purchased only once.
- RFM analysis showed that many customers belonged to Low Engagement or At Risk segments.
- Cohort analysis confirmed weak customer retention across most cohorts.
- The business appeared to rely more on customer acquisition than long-term retention.
Sales & Operations Overview
- KPI cards
- revenue trend
- revenue by state
- top categories
- payment methods
Customer Experience
- delivery performance by state
- lowest rated categories
Customer Retention & Segmentation
- segment distribution
- RFM segment analysis
- retention heatmap
-
Improve logistics in remote regions Delivery delays have a strong negative effect on review scores, especially in regions with longer delivery times.
-
Strengthen customer retention efforts Since repeat purchase behavior is very weak, loyalty programs, personalized recommendations, and post-purchase campaigns could improve customer lifetime value.
-
Monitor high-volume sellers with lower ratings Sellers with strong revenue but weaker customer satisfaction should be reviewed for operational issues.
-
Optimize bulky product categories Furniture-related categories showed lower satisfaction and longer delivery times, suggesting opportunities to improve packaging, shipping, and supplier coordination.
-
Reduce geographic concentration risk Expanding marketing efforts in underrepresented states could support more balanced long-term growth.
-
Prepare for seasonal demand peaks The November revenue spike suggests that Black Friday and holiday campaigns require strong inventory and logistics planning.
Nataliia Butenko
Completed: May 2026