This repository contains a project focused on cleaning and transforming a messy Superstore Sales Data dataset. The dataset includes sales records from a retail business and contains issues such as missing values, duplicates, and inconsistent formatting. Additionally, after cleaning the data, visualisations were created to extract business insights.
The objective of this project is to:
- Clean and preprocess raw sales data to make it analysis-ready.
- Identify and remove duplicates to ensure data integrity.
- Handle missing values using appropriate strategies.
- Standardise formats for dates and categorical variables.
- Create a sample of data visualisations to extract insights for business decision-making.
The dataset used in this project is the Superstore Sales Data, which contains key information about sales, customers, and product categories. Below are the containing columns:
- Row ID: A unique identifier for each row.
- Order ID: A unique identifier for each order.
- Order Date: The date on which the order was placed.
- Ship Date: The date on which the order was shipped.
- Ship Mode: The shipping method used for the order.
- Customer ID: A unique identifier for each customer.
- Customer Name: The name of the customer who placed the order.
- Segment: The customer segment (Consumer, Corporate and Home Office).
- Country: The country where the customer is located.
- City: The city where the customer is located.
- State: The state of the customer.
- Postal Code: The postal code of the customer's location.
- Region: The geographic region where the order was placed.
- Product ID: A unique identifier for each product.
- Category: The category of the product.
- Sub-Category: The sub-category of the product.
- Product Name: The name of the product.
- Sales: The total sales amount for the order.
- Quantity: The quantity of items in the order.
- Discount: The discount applied to the order.
- Profit: The profit made from the order.
The following steps were undertaken to clean the dataset:
- Duplicate Removal: Identified and removed any duplicate records.
- Handling Missing Values: Imputed or removed missing values based on relevance.
- Standardisation: Cleaned column names, reformatted date columns, and ensured consistency in categorical variables.
- Outlier Detection & Treatment: Checked for extreme values in sales and profit columns.
- Feature Engineering: Added new columns:
- Product ID Dup: Identifies duplicated product IDs.
- Updated Product ID: Adjusted identifier for product updates.
- Shipping Duration: Number of days between order and shipping dates.
- Data Types: Ensured proper data types (e.g., numerical, categorical, dates) for each column.
After cleaning the data, visualisations were created to extract key business insights.
This includes:
📊 Metrics Distributions: Distribution of key metrics.
📌 Profit Distribution: Understanding profitability by category and region.
📉 Discount vs. Profit Analysis: Evaluating the impact of discounts on profit.
For a more detailed analysis and conclusions, check out the Data Visualisation Notebook.
- Python (Pandas, NumPy, Matplotlib, Seaborn)
- Jupyter Notebook (for interactive data analysis)
- Git & GitHub (for version control and collaboration)
- File Formats:
- CSV: Standard data storage format.
- JSON: Used for metadata storage.
- Parquet: Optimised for large-scale data processing.
- Pickle: For storing Python objects efficiently.
📂 data/ → Contains raw, processed, and cleaned data.
📂 notebooks/ → Jupyter Notebooks for data exploration, cleaning, and visualisation.
📂 scripts/ → Python scripts for data processing automation.
📂 superstore_sales/ → Configuration and reusable functions.
Key Notebooks:
✔ Data is fully cleaned and formatted for analysis.
✔ Duplicate records were removed, and missing values were handled.
✔ Standardised date formats and categorical data.
✔ Business insights extracted through visualisations.
The dataset is now cleaned, structured, and well-prepared for advanced analysis or dashboard creation. This thorough data cleaning process ensures that the dataset is reliable and accurate, offering a solid foundation for data-driven decision-making and future analyses. It’s a powerful reminder of how critical data cleaning is to ensure meaningful insights and impactful results.
This project is licensed under the MIT License - see the LICENSE file for details.