X5 Group launches AB testing platform

2021: Launch AB Testing Platform

On December 20, 2021, X5 Group announced the launch of the AB testing platform. The system allows you to automatically collect the results of experiments at 18656 stores, as well as analyze the effectiveness of operational processes in all business units of the group.

The AB testing system helps make decisions based on "raw": in data the daily mode information , purchase information is systematized, analyzed and used to draw conclusions regarding each of the hundreds of hypotheses put forward. The main goal of the project is to accelerate the growth of the X5 business by increasing the number of experiments. The system operates on the entire scale of data from the X5 repository - tens of millions of records daily. As of December 2021, the platform is undergoing the testing and revision stage of a number of modules. The planned volume of experiments is more than 1000 per year.

In the development strategy of the X5 Group, one of the key management principles is the data-driven approach - data-based decision-making. Every year, big data experts conduct hundreds of AdHoc of tests. We give our business units a convenient and quick tool for self-testing hypotheses. The own AB platform due to the automation of the process will quickly evaluate the results of the project pilot, accelerate the Time to Market indicator. The implementation of the platform means even deeper integration of analytics tools into the day-to-day processes of the company and improving the accuracy of results by eliminating the human factor, "said Anton Mironenkov, managing director of X5 Technology.

As part of the pilot stage, the AB testing platform was tested by the Pyaterochka retail network to test a number of projects: expanding the range of the Ready Cooking category, the effectiveness of new refrigeration units, the impact of installing bakeries on RTO, etc. The platform's capabilities made it possible to calculate the financial effect of experiments, category growth and other economic indicators in a minimum period of time and without human participation, which made it possible to decide on the quick implementation or closure of the project.

As part of the testing of each hypothesis, stores within these geographical limits are evenly divided into two groups. In the first group, the necessary indicators are accumulated in the system as part of the experiment. In the control group, standard operating activities continue unchanged. After completion of the pilot, the relevant metrics are measured and the success of the pilot is determined.