This results in dissatisfied customers, decreased profits, high holding costs, expired inventory, and inevitable write-offs. But, who bears the responsibility for this inventory mismanagement? Products don't magically appear on shelves, right? They go through a complex distribution process involving multiple middlemen. In a typical One Level Channel distribution, the retailer stands between the manufacturer and the customer, serving as the crucial link responsible for managing inventory to meet customer demand and maximize profit.

Yet, managing inventory is far from simple. It presents countless challenges, including poor demand forecasting, the impact of seasonality and trends, delays in purchasing, limitations in production capacity, delayed fulfillment, and more. Therefore, this work implements a framework to optimize inventory management by identifying the ideal purchase time to minimize overstocking or stockouts while maximizing revenue.

This project tackled the challenge of overstocking and understocking in retail stores. A Machine Learning model was designed to predict ideal purchase timing and quantities, preventing these issues. This solution empowers retailers to:

• Forecast demands a week in advance: This allows for strategic purchase planning, considering vendor fulfillment time.

• Optimize inventory management: By understanding product categories based on profit and demand (cash cows, laggards, commodities, luxury items), retailers can prioritize high-demand, high-profit items (cash cows) while minimizing stock of low-demand, low-profit items (laggards). This ensures they have the right products available to meet customer needs and maximize profits.

Three crucial factors were identified for calculating ideal purchase quantities and timing:

1. Inventory Level: Existing stock helps determine how much to buy.

2. Demand Forecast: The model predicts demand for the next week, ensuring enough products are available.

3. Fulfillment Time: Vendor lead time is factored in to avoid stockouts.

Using these factors, machine learning models were built to forecast demand and calculate potential overstock/understock, and they were compared to the actual inventory levels. 

Core Concepts 

• Dashboarding and EDA: To analyze the basic statistics and generate insights from the data.

• Clustering: To categorize the products based on demand and profitability, which aided in prioritizing products during stock planning.

• Association: To identify frequently bought product types and flavor combinations to understand customer preferences.

• Regression: To predict product demand weekly using demand per week, helping in the allocation of purchase quantities.

Tools and Algorithms Used

• Python, Power BI, SQL, Postgres Database, Excel, and UML 

• K-Means Clustering, Hierarchical Clustering, Apriori, NaiveBayers, Decision Tree, Support Vector Machine