CafeZupas: Smart ingredient forecasting—enhancing inventory accuracy with dish-level demand prediction
Project Overview
A demand forecasting model was developed for CafeZupas to predict the quantity of raw ingredients required for dish preparation. The solution aimed to optimize inventory management, reduce food waste, and ensure the restaurant could consistently meet customer demand. By integrating this model with the restaurant’s POS system, it provided real-time data input to enhance accuracy.
Problem Statement
Previous demand forecasts at CafeZupas led to stockouts, where ingredients were unavailable, and over-purchasing, resulting in excess stock and food waste. The high food waste increased operational costs and affected profitability. Additionally, the lack of automated insights in the supply chain planning process caused inefficiencies, making it difficult to optimize inventory management.
Key Findings
- Peak Sales Periods: Analysis revealed recurring demand spikes—particularly over weekends—which guided more efficient ordering cycles and preparation plans.
- Seasonality in Ingredients: Certain ingredients showed strong seasonal dependencies, prompting the development of custom models to better predict their usage across the year.
- Impact of Menu Changes: Menu updates significantly influenced ingredient demand, underscoring the importance of flexible forecasting models that adapt to evolving customer preferences.
Implemented Solution
Developed dish-level forecasting models using ARIMA and Prophet, integrated with POS for real-time data, and created a dashboard for trend monitoring and inventory adjustments:
-
Dish-Level Forecasting Models:
Developed individual forecasting models using ARIMA and Facebook Prophet for each menu item, accounting for unique ingredient compositions and sales history.
-
Cross-Validation Framework:
Introduced robust cross-validation techniques to minimise forecast errors and ensure model reliability over time.
-
Seasonality & Trend Integration:
Embedded seasonal and trend-based features to adjust for fluctuations in ingredient usage, improving the alignment between forecasts and actual sales.
-
POS System Integration:
Connected the models with CafeZupas’ POS system to allow real-time data streaming, improving responsiveness and predictive accuracy.
-
Monitoring Dashboard:
Built an intuitive dashboard for inventory managers to visualise trends, track forecast accuracy, and make timely adjustments to stock levels.
Results
The forecasting solution reduced food waste by 10% and improved forecasting accuracy by 15%, ensuring that ingredients were available when needed and reducing surplus. This led to more consistent meal preparation and fewer disruptions in kitchen operations. The system also improved supplier coordination, with a 20% boost in relationship efficiency through timely, data-backed orders. Overall, the model helped CafeZupas create a more sustainable and cost-effective approach to daily meal prep and inventory management—while enhancing service reliability for customers.