Sales Forecasting Dashboard — Decision-Focused Time Series Analysis

Problem

Retail teams often struggle to anticipate sales fluctuations, leading to stockouts, overstocking, and inefficient resource allocation.
Traditional dashboards show historical trends, but rarely provide actionable forecasts for decision-making under uncertainty.

---

What this project demonstrates

This project implements a time series forecasting framework to generate interpretable, actionable predictions that inform inventory, marketing, and financial planning decisions.

The focus is on decision support, not just visualizations.

---

Decisions this system supports

• Forecasting monthly sales to optimize inventory levels
• Planning staffing and resource allocation based on expected demand
• Timing marketing campaigns for peak revenue months
• Estimating revenue and detecting potential shortfalls
• Understanding seasonality and growth trends for strategic planning

---

Analytical approach

Data

• Historical retail sales: 2015–2018
• Aggregated monthly by region and product category
• Features: sales amount, product category, region, customer demographics

Forecasting methods

• Holt-Winters Exponential Smoothing (trend + seasonality)
• Comparative evaluation against Linear Regression and Moving Average
• Forecast uncertainty quantified with confidence intervals
• Optional: additional methods (Prophet, ARIMA) for model comparison

Model selection

• Holt-Winters selected due to superior predictive performance (MAE $13,133 vs $34,823 baseline)
• Captures seasonal peaks and growth trends accurately
• Model evaluation emphasizes decision relevance over raw accuracy

---

Example: Forecast outcome

Scenario: Total monthly sales for 2019

Predictions

• Total Sales: $899,039
• Expected Growth: +24.5% vs 2018
• Peak months: November ($114K), December ($116K)

Business interpretation

• Inventory planning: stock levels adjusted for seasonal peaks
• Marketing: campaigns aligned with predicted high-demand months
• Risk management: early warning for potential revenue shortfalls

---

System overview

• Data ingestion: CSV, aggregated time series
• Forecasting engine: Python, Statsmodels
• Visualization: Streamlit + Plotly
• Outputs: Forecast tables, interactive charts, downloadable CSVs

---

Architecture (compressed)

• Backend/Engine: Python 3.8+, Statsmodels, NumPy, Pandas
• Dashboard: Streamlit + Plotly for interactive visualization
• Evaluation: MAE, RMSE, R² score comparison across models

---

Why this matters for Data Science

Forecasting is decision-oriented modeling under uncertainty.
This project demonstrates:

• Time series decomposition and trend/seasonality detection
• Model evaluation for actionable outcomes
• Translation of analytical results into business decisions
• Understanding uncertainty and risk in forecasts

---

Project structure

sales-forecasting-dashboard/
│
├── data/
│   └── SuperstoreSalesDataset.csv
│
├── notebooks/
│   └── 01_sales_forecasting_analysis.ipynb
│
├── models/
│   └── sales_forecast_model.pkl
│
│── screenshots/
│     └── forecast.png
│     
├── app.py
├── requirements.txt
├── README.md
└── .gitignore

Getting Started

Prerequisites

• Python 3.8 or higher
• pip package manager

Installation

1. Clone the repository

git clone https://github.com/Emart29/sales-forecasting-dashboard.git
cd sales-forecasting-dashboard

1. Install dependencies

pip install -r requirements.txt

1. Run the application

streamlit run app.py

1. Open your browser and navigate to http://localhost:8501

Screenshots

Forecast Visualization

Author

[Emmanuel Nwanguma]

• LinkedIn: Nwanguma Emmanuel
• GitHub: Emart29
• Email: [email protected]