Financial KPI Forecasting Dashboard

A real-world time series forecasting projec

Dashboard Preview

Live Demo: [https://ecommerce-kpi-forecasting-mvhym4nxtvdktypufiaipf.streamlit.app/]

Video Demo: [https://youtu.be/kAJ2TPjXUhU]

Local Access: streamlit run app/streamlit_app.py

Main Features:

• Visualize KPI trends over time
• 6-month forecast using Prophet
• Summary metrics like growth rate vs past average
• Table of future values with download/export potentialt key performance indicators (KPIs) for a retail e-commerce business. This interactive dashboard was built using Prophet, Pandas, and Streamlit, designed to assist business stakeholders in making data-driven decisions by forecasting sales, quantity, and order volumes.

---

Project Overview

In today's competitive business landscape, accurate forecasting of financial metrics is critical for demand planning, inventory control, and revenue optimization. This project builds a forecasting system using real transactional data from an online retailer and presents it through a user-friendly dashboard.

---

Key Objectives

• Clean and transform raw retail sales data into meaningful KPIs
• Use Prophet time series models to forecast:
  • Total Revenue
  • Total Quantity Sold
  • Total Orders
• Present historical trends and future projections in an interactive dashboard using Streamlit
• Empower users to interpret and compare forecast vs actual KPI performance

---

Project Structure

ecommerce-kpi-forecasting
├── app/
│   └── streamlit_app.py         # Streamlit dashboard code
├── data/
│   ├── ecommerce_data.csv       # Original cleaned dataset
│   ├── monthly_kpis.csv         # Aggregated monthly KPIs
│   ├── revenue_forecast.csv     # Prophet forecast for revenue
│   ├── quantity_forecast.csv    # Prophet forecast for quantity
│   ├── orders_forecast.csv      # Prophet forecast for orders
│   ├── arima_revenue_forecast.csv
│   ├── orders_arima_forecast.csv
│   └── quantity_arima_forecast.csv
├── models/                      # Saved model files
├── notebooks/
│   └── 01_data_cleaning_eda.ipynb  # Data cleaning and KPI calculation
├── requirements.txt             # Python dependencies
└── README.md                    # Project documentation

---

Forecasting Approach

We use Facebook Prophet, a robust and easy-to-use time series forecasting library, ideal for business data. It automatically handles:

• Trends
• Seasonality
• Holiday effects
• Missing values and outliers

Prophet is especially powerful for data with strong seasonal patterns — like sales.

---

KPIs Forecasted

KPI	Description
Total Revenue	Total sales revenue per month
Total Quantity	Number of units sold
Total Orders	Number of unique invoices per month

All KPIs are derived from real e-commerce transactions.

---

Tools & Technologies

Category	Stack
Data Handling	Pandas, NumPy
Time Series Model	Facebook Prophet
Dashboard	Streamlit, Plotly
Forecast Evaluation	sklearn.metrics (MSE)

---

Dashboard Preview

Live Demo: https://ecommerce-kpi-forecasting-mvhym4nxtvdktypufiaipf.streamlit.app/

Local Access: streamlit run app/streamlit_app.py

Main Features:

• Visualize KPI trends over time
• 6-month forecast using Prophet
• Summary metrics like growth rate vs past average
• Table of future values with download/export potential

---

How to Run Locally

1. Clone the Repo

git clone https://github.com/SachinthyaCodes/ecommerce-kpi-forecasting.git
cd ecommerce-kpi-forecasting

2. Install Dependencies

pip install -r requirements.txt

If prophet fails on Windows, try:

pip install prophet --force-reinstall --no-binary :all:

3. Run the App

streamlit run app/streamlit_app.py

---

Results Summary

KPI	Model Used	MSE (last 6 months)	Notes
Total Revenue	Prophet	Lower than ARIMA	Prophet performed better
Total Quantity	Prophet	Accurate and stable
Total Orders	Prophet	Stable forecast

ARIMA was tested but Prophet was selected for its superior accuracy, flexibility, and usability.

---

Key Learnings

• Handling real-world messy transactional data
• Building robust forecasting pipelines with Prophet
• Creating end-to-end, deployable dashboards with Streamlit
• Evaluating models with real KPIs like MSE
• Translating technical insights into business-friendly visuals.

---

Author

Sachinthya Lakshitha
Reach me on LinkedIn or sachinthyaofficial@gmail.com