Sales-Forecasting-for-Inventory-Optimisation

📈 Sales Forecasting for Inventory Optimization

---

---

📘 Overview

This project focuses on improving inventory management for an independent supermarket through accurate sales forecasting. Using Excel-based time series modeling, it compares multiple forecasting techniques—Holt’s Linear Trend, Holt-Winters Seasonal Model, and Damped-Trend Exponential Smoothing (SES with damping)—to determine the most effective method for minimizing stockouts and overstock.

---

🎯 Objectives

• Forecast monthly sales using time series methods
• Evaluate methods using error metrics (MAE, MAPE, RMSE)
• Recommend a forecasting strategy to optimize stock ordering
• Ensure replicability using only Excel and Solver

---

🧾 Dataset

• Source: Daily sales volume (Excel)
• Time Range: Jan 1, 2009 – Oct 24, 2020 (hypothetical)
• Entries: 4,315 observations
• Aggregation: Daily to monthly totals for clearer trend/seasonality

---

📊 Methods Applied

🔹 Holt’s Linear Trend (Method A)

• Captures trend without seasonality
• Uses Excel Solver to optimize level & trend parameters

🖼️ Fig A: Solver Output – Holt’s Linear Trend

---

🔹 Holt-Winters Seasonal (Method B)

• Captures both trend and seasonality
• Optimizes alpha, beta, and gamma via Solver

🖼️ Fig B: Solver Output – Holt-Winters Seasonal

---

🔹 Damped-Trend Exponential Smoothing (Method C)

• Adjusts for fluctuating trends using a damping factor
• Solver minimizes absolute error and RMSE

🖼️ Fig C: Solver Output – Damped Trend ES

---

🔹 Combined Forecast (Method D)

• Averages Methods A, B, and C for improved accuracy
• Leverages the strengths of each individual model

🖼️ Fig D: Combined Forecast Output

---

📈 Evaluation: Error Metrics

Method	MAE	MAPE	RMSE
A	82,260	25.4%	18,192.65
B	82,468	25.6%	18,219.33
C	82,937	26.4%	18,174.48
D	69,740	21.09%	16,854.37

🖼️ Fig E: Forecast Accuracy Comparison

---

✅ Findings & Recommendations

• The Combined Forecast (Method D) is the most accurate and reliable.
• Use Method D for monthly order decisions to minimize stockouts and overstock.
• Monitor forecast vs. actual sales monthly to recalibrate model if needed.
• Solver settings and parameters are fully documented in the workbook.

---

🔁 Replication & Audit

• Each method is clearly structured in Excel with labeled parameter cells
• Solver steps are reproducible
• “Cleaned Data” tab shows aggregated input with original trend patterns

---

🚀 Getting Started

1. Download the Excel workbook from this repository
2. Open Enterprise Data.xlsx Get Dataset Here
3. Navigate to each method’s worksheet
4. Use Solver to minimize RMSE and observe forecast outputs

---

📚 Reference

Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2020).
The M4 Competition: 100,000 time series and 61 forecasting methods.
International Journal of Forecasting, 36(1), 54–74.
https://doi.org/10.1016/j.ijforecast.2019.04.014

---

👤 Author

Ramanav Bezborah
🔗 GitHub Profile