Fix Price has implemented a communications quality control system

2024

Implement audio analytics to assess the quality of cashiers' interactions with customers

Fix Price has launched a project to introduce audio analytics to assess the quality of cashiers' interaction with customers. In the fourth quarter of 2024, the solution was implemented in 20 stores managed by the company in Russia, and in the first quarter of 2025, coverage will expand to 200 such points. The results of the pilot project will be summed up by mid-2025. The company announced this on January 21, 2025.

The technology records audio data from cash desks and transfers it to the data center, where audio files are converted into text using language models. Based on the analysis of texts, the system assesses the key aspects of interaction with buyers: courtesy in service, offering a loyalty card, offering a package, as well as identifying possible conflict situations. When such situations are identified, the store manager is automatically notified for prompt response.

{{quote 'author=said Oleg Lexin, Chief information officer of Fix Price.|We consistently implement solutions to improve the quality of service in our stores. Previously, we implemented video analytics projects to monitor queues, "empty shelves" and the condition of trading floors. Together with Smart Retail Tech, we have implemented audio analytics and are testing it to assess the quality of cashiers' interaction with customers. The first results showed that the solution accurately records compliance with communication standards and effectively identifies conflict situations. This allows us not only to assess the quality of service, but also to work on its continuous improvement,

}}

Introduction of AI into the process of processing internal applications in the company

The Fix Price network has launched a pilot project based on artificial intelligence (AI), the purpose of which is to categorize internal applications in the company for subsequent processing by support team specialists. The company announced this on March 20, 2024. This allowed to automate the process of analyzing primary requests to IT support, which led to a reduction in the time for transferring a case and finding a specific performer. The next step in the development of the project will be to automate the process of analyzing requests to other divisions of the company.

Since the launch of the service, about 100 thousand applications have already been processed.

Every year, the company's IT division receives more than 200 thousand requests for solving various tasks and issues from employees. Despite the proposed categorization of tasks, many of them fall not to specialized specialists or the wrong category, which increases the waiting time for an answer and slows down processes. Thanks to the implemented service, requests are now analyzed by AI. A significant amount of accumulated structured data allowed us to successfully launch a pilot project, train the necessary AI models and get intermediate results for recognizing applications at a level above 85%, "said Oleg Lexin, head of the Fix Price IT service.

To solve the problem, a set of natural language processing methods was used, including training, further training and integration of an ensemble of various catboost machine learning models (open source library with the implementation of the gradient boosting algorithm from Yandex), rubert (open-source distilled model implemented on the basis of BERT from Google) and a number of auxiliary open-source technologies.