Dodo Brands Implements Intelligent Ingredient Flow Prediction System

2021: Implementation of Azure Machine Learning-based Ingredient Flow Prediction System

Dodo Brands, in a technology partnership with Crayon and Microsoft, has implemented an ingredient flow prediction system based on Microsoft's Azure Machine Learning. This was reported on March 1, 2021 by Microsoft Rus (Microsoft Rus). It is expected that the implementation will allow the company to save up to 54 million rubles a year. As of March 1, the system is already operating in 50 restaurants of the company.

Project prerequisites

The Dodo Pizza chain is part of Dodo Brands and operates in Russia and Kazakhstan. The success of pizzerias depends on many factors. The key ones include properly planning and ordering the ingredients you need.

Previously, each pizzeria independently determined the volume of purchases of products. The managers made the calculation of the required number of ingredients in simple tables, or formed an order approximately. Due to the lack of a single algorithm and software, managers spent up to 4-5 hours a week on calculations and placing orders. The calculation of the required amount of ingredients was made in simple tables or formed approximately. There were often situations with both re-procurement and a lack of ingredients. In addition, difficulties arose due to mail and Internet failures, lack of experience among employees of new pizzerias, errors due to the human factor.

Errors with ordering ingredients worsened customer experience and reduced revenue. Dodo Pizza needed a centralized system that would make it easier to predict the consumption of ingredients based on the demand for menu items in each specific restaurant.

Decision

The solution was the development of an intelligent predictive system that takes into account factors such as various trends in data (for example, demand growth as spring warms), the holiday calendar, company marketing activities, local events and many others. The system is integrated with the Dodo IS platform, which combines tools to manage all restaurant business processes in a single interface.

Project results

According to Dodo Pizza, now the restaurant manager can get a list of ingredients necessary for ordering in 2 minutes. The implementation made it possible to reduce the workload of the restaurant team, as well as increase the accuracy of forecasting the purchase of ingredients by 18%. The total savings from scaling the Dodo Pizza solution is estimated at up to 4.5 million rubles per month, given the reduction in direct revenue losses and a decrease in the number of urgent deliveries required in case of a lack of ordered ingredients. Thus, the company plans to save up to 54 million rubles annually.

"The automated process of calculating the order reduces the labor costs of managing our pizzerias (on a network-wide basis, this is a savings of up to 1.1 million rubles per month), and also reduces the burden on the logistics system. Most importantly, the use of the algorithm increases the accuracy of forecasting, which means it reduces revenue losses, "said Evgeny Leontyev, Head of Logistics, Dodo Brands.

The system was developed by Crayon, a Microsoft partner, in collaboration with the IT Dodo Brands team based on a solution Microsoft Azure ML that was used to analyze data, select frameworks and train test models. The Azure Databricks service was also used to speed up calculations.

"The task was interesting: after analyzing the data, it became clear that there are a fairly large number of different classes of ingredients whose consumption patterns differ from each other. For each of these classes, a separate model had to be selected for prediction. I also want to note that we were pleasantly surprised by the level of readiness of Dodo to solve this problem: when we started solving the problem, we were quickly and clearly provided with all the necessary data, infrastructure in the Microsoft Azure cloud and provided all the necessary integrations. Due to such a mature approach, we managed to quickly implement the project, in fact, all the work was completed by us in a week and a half, "added Vladimir Eronin, director of the competence center for artificial intelligence and data analysis, Crayon Russia.

"The willingness to quickly meet market demands and implement progressive technologies allows Dodo Brands to maintain high growth rates for many years. Our joint case is a great confirmation of this. The high level of technological expertise of our partners, Dodo Brands and Crayon, made it possible to achieve the implementation of the project and results in such a short time, "said Vladislav Shershulsky, Head of Advanced Technologies at Microsoft in Russia.

Project Development Plans

The next phase of the project will include further deployment of the system to the restaurant chain, as well as its integration with ingredient supplier systems. Auto-billing can be scaled and applied to all Dodo Brands countries and concepts. The companies also plan to jointly develop other forecasting models, for example, to launch new products, implement marketing campaigns, plan local events, etc.