Marketing analytics for BelkaCar carsharing service

BelkaCar, a car-sharing service operating in Russia since 2016, has about 6,000 vehicles and over 1.5 million users in Moscow and Sochi. With a growing customer base, the service needed a 'smart' program—a tool for managing and optimizing demand for vehicles. To enhance service quality, the BelkaCar team, in collaboration with Epoch8, began working on this solution.

We gathered significant anonymized data in a single location (a DWH on Google BigQuery), cleaned it, and combined it into key analytical entities:

A Particularly Interesting Task — Structuring Trip Data.

Each trip has starting and ending coordinates, and to analyze such geodata, BelkaCar uses an approach developed by Uber engineers. The entire service area is divided into a hexagonal grid (H3 cells). Each cell is assigned metadata, including information about demand, supply, and trip events.

The maps above illustrate the process of grouping points using H3: cars in the city; cars within hexagons; hexagons shaded based on the number of cars.

We develop a Customer Data Platform (CDP) — a repository for collecting and storing customer data.

All collected data is loaded into the CDP and regularly updated. As a result, we created a platform that contains a large set of structured data, which can be used to generate analytical reports and run machine learning models.

All data from the CDP is displayed on dashboards, ensuring the business always knows the status of all users and vehicles.

Marketers can analyze user behavior by using graphs and, based on this information, develop and implement marketing campaigns.

For example, the business may face the task of balancing supply and demand for cars in a specific location.

When a potential driver opens the app, their request goes through the CDP. At this point, a machine learning model predicts where the user is likely to travel.

If our forecast indicates that the driver will complete their trip in a high-demand location, the price will be made more attractive to increase the likelihood of them using the service.

This approach encourages the user to move a car from a low-demand area to a high-demand area, reducing expected idle time as a result.

The Customer Data Platform (CDP) allows you to identify all users who meet specific criteria.

With the CDP, not only can ML models segment users, but the marketing team can also create segments for push notifications or email campaigns based on specific characteristics.

With over 100 parameters available, it’s possible to accurately define segments and work with them effectively.

Thanks to the CDP, customers can: