M.Video-Eldorado forms an assortment on shelves in stores using big data

2021: Create an assortment on shelves in stores using big data

The M.Video-Eldorado Group announced on August 17, 2021 that it had automated the planning of the assortment available in stores based on depth analytics. Artificial intelligence, analyzing the behavior of customers in choosing products, forms the optimal set of models on shelves in stores to meet the needs of different groups of customers. According to the results of the pilot in more than 100 cities, the machine algorithm, simply changing the assortment on the shelves, allowed to increase sales in test categories to 3.5%.

The M.Video-Eldorado Group has developed and launched MVP solutions (minimum viable product or minimum viable product) for automated assortment planning on shelves in stores. In the context of the development of the hybrid model, the retailer analyzes the strategy of choosing customers online and uses this data to improve the customer experience in retail stores. In-depth analytics technology uses sales and customer behavior data - search sessions, views, and comparisons, and takes into account specified business targets. After processing all the information, the machine assistant recommends to commercial managers working with the assortment what articles to put on the shelf in each department of the store in a limited space.

For example, the user watches and compares multiple TVs with the same diagonal, screen resolution, and functionality (HDMI connector, voice control, or Wi-Fi connection). It can be assumed that these goods close its specific need. The task is to identify and try to meet these needs, not cluttering the store with similar goods, but making the shelf attractive to a large number of customers, closing the maximum number of needs.

Шаблон:Quote 'author = said Oleg Muravyov, Commercial Director of M.Video-Eldorado Group.

As part of the pilot, the model independently formed assortment matrices of several categories: headphones, kettles, washing machines, and proved its effectiveness. Using the offline A/B testing methodology developed at M.Video, a significant increase in sales in pilot categories up to 3.5% was determined compared to comparable stores of the control group. The company has begun scaling assortment planning and is already forming a third of the assortment of stores on the basis of a statistical algorithm. By the end of 2021, the retailer plans to fully automate this process.

Automatic assortment planning consists of two stages. On the first tree , client needs are formed. The digital solution analyzes user sessions and clusters products, identifying interchangeable models that are as close as possible in comparison and viewing, in other words. Then the selected requirements are loaded into the optimizer - a machine algorithm for optimizing the assortment for store shelves. It generates a list of goods for each category and demand, which maximizes business indicators in terms of turnover, margin and number of checks, and also takes into account the index of uniqueness of goods within the store matrix. All parameters can be customized and adapted quickly as market conditions change and the category strategy evolves. Moreover, the optimizer takes into account the size and specificity of demand in different stores - about 20 categories of stores are included in the model, for each of them the algorithm recommends a separate product filling.

Шаблон:Quote 'author = noted the head of product management "Assortment Planning" M.Video-Eldorado Maxim Nikolaev.