The Group has created a Data Science Center that develops solutions based on data analysis and machine learning algorithms in numerous areas: communicating with customers, creating personalized offers and working with the product range and promotional plan. In 2020, the center’s team more than tripled in size, from 20 to 75 people.
Creating predictive models for personalization in the consumer electronics segment requires extensive research and experimentation, as well as innovative approaches to working with data and algorithms due to the wide range and low frequency of purchases. Nevertheless, the team at M.Video–Eldorado’s Data Science Center has successfully introduced models that can:
Using these models across various customer touchpoints (marketing campaigns, the mobile app, website and retail) enables the Group to personalize the shopping experience and interaction, as well as increase conversion rates and the average ticket.
In mid–2020, the Data Science Center began focusing on ways to develop speech analytics. Analytical models are primarily applied by creating chatbots that automate standard customer queries, thereby reducing the load on call center operators and boosting service speed.
In addition to chatbots, models have been introduced in speech analytics to analyze customer product reviews and highlight key features from them.
Analyzing customer reviews and queries helps to identify important product features for use in assortment planning, as well as improve scenarios for recommendations and product selection.
Machine learning algorithms help to make demand forecasting for certain items more accurate and take into account correlations in data that are not obvious to humans.
As such, the introduction of machine learning algorithms can significantly optimize not only forecasting itself but also associated costs, such as for the use of storage facilities or transport.
In 2020, the Data Science Center team was one of the first in the market to use cloud technologies to manage machine learning infrastructure by creating environments in the cloud for the development and use of their products. By transitioning to the cloud, we managed to triple the amount of resources for machine learning development, which made it possible to simultaneously conduct around 100 pilot projects and experiments, while also significantly reducing costs. In 2020, the cloud was used for machine learning tasks. In 2021, we plan to transfer analytical data storage to the cloud in an effort to gain even greater flexibility and further growth opportunities.
