Sales Data Analysis Dashboard

A production-ready analytics project built with Python and Streamlit. It covers data cleaning, EDA, interactive BI dashboards, forecasting, anomaly alerts, and executive reporting.

Project Overview

The project can analyze:

• Default dataset (data/amazon.csv -> processed to retail format)
• Uploaded files (.csv, .xlsx, .xls)
• SQL database sources (connection URL + table/query)

All inputs are normalized into a common analytics schema before visualization and forecasting.

Features

• Data pipeline

  • Data cleaning, deduplication, type conversion
  • Schema normalization for common sales datasets
  • Unified fields: Order Date, Sales, Profit, Category, Sub-Category, Region, Customer Segment

• Interactive dashboard (Streamlit + Plotly)

  • KPI cards and dynamic filters
  • Interactive trend, category, region, and correlation views
  • Dataset Info tab (raw + processed transparency)

• Forecasting (real-world evaluation)

  • Multi-model comparison: Linear Trend, Seasonal Naive, Exponential Smoothing
  • Single holdout benchmark and rolling-origin backtest
  • Auto-selection of best model by error metrics

• Alerts and actions

  • Monthly anomaly detection (z-score)
  • Low-margin category detection
  • Negative-profit transaction detection
  • Region risk summary
  • User-adjustable thresholds from sidebar

• Data quality and drift monitoring

  • Rule-based quality checks (required columns, minimum rows, missing ratios, duplicates)
  • Quality score and critical/warning failure tracking
  • Drift table (current slice vs baseline) using mean-change and KS statistic
  • Configurable fail-fast quality gates for scheduled jobs

• Executive reporting

  • Download report as CSV and Markdown from UI
  • Save timestamped snapshots to workspace
  • CLI job runner for scheduled report generation with strict quality/drift options

Tech Stack

• Python
• Pandas, NumPy
• Plotly, Streamlit
• Matplotlib, Seaborn (static artifact generation)
• SQLAlchemy (database source)
• Jupyter Notebook

Project Structure

Sales_Data_Analysis/
|-- app.py
|-- analysis.ipynb
|-- requirements.txt
|-- README.md
|-- TECH.md
|-- data/
|   |-- amazon.csv
|   `-- retail_sales.csv
|-- outputs/
|   |-- figures/
|   `-- reports/
`-- src/
    |-- anomalies.py
    |-- config.py
    |-- data_processing.py
    |-- eda.py
    |-- forecasting.py
    |-- insights.py
    |-- jobs.py
    |-- pipeline.py
    |-- quality.py
    `-- reporting.py

Documentation

For detailed technical documentation, architecture diagrams, design patterns, and API reference, see TECH.md.

Key Documentation Topics:

• System Architecture - High-level design and data flow
• Core Modules - Detailed documentation of each component:
  • Data Processing Pipeline
  • Quality Assessment & Drift Monitoring
  • Anomaly Detection Engine
  • Multi-Model Forecasting
  • Executive Reporting
  • CLI Job Runner
• Data Schema - Unified analytics schema definition
• Technology Stack - Complete dependency list
• Deployment Guide - Production deployment options
• Extension Points - How to add new features
• Troubleshooting - Common issues and solutions

Run Locally

1. Clone the repository

git clone https://github.com/Ayushkumar418/Sales_Data_Analysis.git
cd Sales_Data_Analysis

2. Install dependencies:

pip install -r requirements.txt

3. Build processed dataset + static figure artifacts:

python -m src.pipeline

4. Run dashboard:

streamlit run app.py

Note: If streamlit run app.py doesn't work on your system, try:

python -m streamlit run app.py

Scheduled Reporting

Generate a report snapshot from terminal:

python -m src.jobs --source default --output-dir outputs/reports

Strict quality gate example:

python -m src.jobs --source default --output-dir outputs/reports --strict-quality --fail-on-drift --min-rows 100 --max-missing-pct 0.15 --drift-mean-change-threshold 20 --drift-ks-threshold 0.25

Database-based snapshot:

python -m src.jobs --source database --connection-url "sqlite:///data/sales.db" --sql-query "SELECT * FROM sales" --output-dir outputs/reports

Use these commands with Windows Task Scheduler or cron for recurring automatic report generation.