Case Study08/26/2020

Optimizing restaurant inventory using ingredient-level demand forecasting

Joseph Day
Data Scientist
Jacob Zweig
Co-Founder, Principal Data Scientist

Overview

In a recent project, Strong Analytics partnered with a leading food distributor to develop a platform that could help restaurants improve their operations by maximizing revenue and minimizing food waste. To do so, we helped them implement a cutting-edge machine learning solution that used ingredient-level forecasts based on diners’ orders to anticipate the restaurants’ restocking needs.

The Challenge

The food distributor approached Strong with an open question – how could they provide value to their customers and enable smarter operations?

We collaborated with their internal stakeholders to understand the key challenges a restauranteur faces. It was clear that the process of managing an ingredient inventory was a highly manual and challenging task. Each restaurant had their own internal heuristics that they had developed from experience, but most did not leverage optimal, data-driven approaches.

Creating an optimal ordering strategy to maximize profit and minimize food waste thus provided an opportunity to create substantial improvements.

In the restaurant industry, managing the risk of missed demand with that of food wastage is a careful balance for operators. Deviating too far in either direction can create substantial impacts to profitability. Beyond that, risks vary based on ingredients, dishes, seasons, and more. Some ingredients are present in many dishes and their availability may impact margins substantially. Others can have extremely short shelf life and require frequent restocking. Creating an optimal ordering strategy to maximize profit and minimize food waste thus provided an opportunity to create substantial improvements.

Our Solution

To solve this problem, Strong built a solution that helps restaurant operators create optimal ordering plans with detailed inventory management, forecasting, and visualization.

Strong created a platform to forecast the inventory of each ingredient into the future and inform restauranteurs when they need to replenish inventory and how much they will need to reorder to optimize profit.

Using point-of-sale data in conjunction with order histories and menu items, Strong created a platform that leverages advances in dynamic systems theory to forecast the inventory of each ingredient into the future. Importantly, this also informs when they need to replenish inventory and how much they will need to reorder to optimize profit.

Restauranteurs are then able to automatically generate optimal orders and submit them with two clicks, minimizing the burden of manual inventory management and ordering. This functionality, which balanced the shelf life of ingredients, their supply, relative importance for the restaurant, local and global trends, and more, created restocking strategies that substantially improved restaurant profitability compared to existing strategies.

Beyond creating individual ordering strategies, by aggregating ingredient and dish-level trends across restaurants, we were able to show restaurant owners how consumer’ tastes vary with the seasons and evolve over time. These detailed insights, which were previously not available in any existing tools, provided substantial utility to restaurants and improved their ability to continuously iterate and improve their menu engineering.

Share: