Our client, one of the world’s largest coffeehouse chains, experiences a daily demand of over 500 units at every store across India. The client expressed the desire to predict the daily demand for all coffee products one week in advance for every store. Additionally, they sought the development of an interactive dashboard to provide a centralized solution for monitoring predicted coffee demand across all geographies.
To achieve the client’s objectives, we followed a comprehensive approach:
Data Analysis: We analyzed loyalty and walk-in transactions for each store format within every district. Our study focused on understanding buying patterns and drink preferences for the fiscal year 2022-23.
Feature Engineering: To enhance our predictive models, we created additional features, including seasonality, discount indicators, and holiday flags, based on the loyalty transactions.
Demand Forecasting Models:
- Approach 1: Univariate time series demand forecasting was employed for the Walk-in Customer base. This approach aimed to predict daily demand based on historical trends and walk-in customer behavior.
- Approach 2: For the Loyalty Customer Base, we utilized multivariate demand forecasting techniques. This involved considering various factors that influence demand, such as customer loyalty, past purchasing behavior, and other relevant variables.
Aggregated Predictions: To obtain a holistic view, we multiplied the predictions received at the Store Format level by the contribution factor of each store under the format. This allowed us to generate store-level demand predictions.
The implementation of our forecasting system had a significant impact on our client’s operations:
Accuracy: The client achieved an accuracy rate of 85% in forecasting daily demand. This high level of accuracy provided them with confidence in their inventory management.
Inventory Management: Accurate daily demand predictions enabled the client to prevent stockouts and effectively manage their inventory. This, in turn, resulted in a reduction in inventory costs.
Resource Allocation: The forecasting system empowered the client to perform dynamic resource management at the district level. They could allocate the optimum resources to each store based on the predicted demand, ensuring efficient operations and resource utilization.