oneFactor has launched an ML platform for MegaFon, Skyeng and Avito

2022: Launch of a machine learning platform to optimize key business processes

On September 13, 2022, oneFactor announced the launch of a machine learning platform (ML platform) to optimize key business processes based on processing large amounts of data. With the ML platform, analysts can build models to assess risk, segment the client portfolio, and improve service quality. The first users of the platform were MegaFon, Skyeng and Avito.

In 2022, MegaFon and Skyeng began using oneFactor software, which runs on machine learning technologies. The results of the two months of use of the platform showed a significant increase in the efficiency of business processes, customer satisfaction, as well as significant savings in operating process costs. In addition, during the month, the software was tested in Avito, within the framework of which a model was built to effectively and promptly prevent potentially unsafe registrations based on data connected to the platform.

Skyeng is actively using the oneFactor ML platform to build personalized communications with users and prioritize incoming leads based on payment probability. A model was built on the platform, which was trained on the results of calls according to the company's historical data. The model allows you to predict the success of the call and determines the optimal script for communication with the client.

With the help of the oneFactor ML platform, MegaFon analysts identified cases in the mobile operator's business model that could be significantly optimized by applying big data solutions, for example, predicting the best time to contact a subscriber. The results of the work showed a significant increase in customer satisfaction.

The oneFactor ML platform allows analysts to get a ready-made full cycle of developing machine learning models: collecting internal and external data in automatic mode, training models, putting into operation (MLOps) and AutoML in minutes without the help of IT services, as well as executing models in various modes, including real-time processing, butch and regular monitoring.

For example, to determine the correct time for contact with the client, anonymized data of subscribers of the cellular operator with information about the services used by them are uploaded to the platform. Then the model is trained using algorithms implemented by oneFactor. Also, the model is supplemented with third-party data and in the future the system itself recommends at what particular hour it is better to communicate with the client.

Decisive for us when choosing the oneFactor ML platform for a number of real-time cases was the rapid implementation, the availability of ready-made modeling tools, high performance and the quality of the solution as a whole. We stated high requirements for our project - the execution time of the request in real time is 500ms with an average load of at least 300RPS (Requests Per Second). The oneFactor platform allows you to withstand such high speed and performance values ​ ​ of ML models due to the easy horizontal scaling capability embedded in the architecture. This allowed us to make MegaFon's service even more convenient for customers, "said Sergey Fedorchenko, Director of Enterprise Data Management at MegaFon.

Despite having our own team and inhouse infrastructure for machine learning and building data driven processes, we are actively using the oneFactor ML platform. This helps us in the processes of attracting new customers, managing marketing campaigns and building more relevant personalized processes for each user of our products, - comments Irina Bubnova, Skyeng sales manager.

Big data-based services are an important source of additional product value and a tool for solving business problems, including in the framework of working with user security. It is important for us to work proactively, preventing violators from accessing the Avito platform. An algorithm based on machine intelligence analyzes up to 8 billion clicks every day, identifying potentially unsafe user actions and identifying invisible links between accounts. Cooperation with oneFactor will help us continue to develop smart algorithms and remain one of the safest Internet platforms on the market, "said Andrei Rybintsev, director of the Trust & Safety department.

According to oneFactor CEO Stanislav Murin, "big data products are one of the key drivers of growth in the Russian economy. For us, one of the key tasks of the company is to provide our partners with high-quality tools and expertise to solve business problems using machine learning. And we are glad that our clients use domestic high-tech developments to solve their tasks and those of their users. "

One of the competencies of oneFactor is the creation of products based on a machine learning platform with a short time-to-market. The trained model can be used in operation as early as 30 minutes after publication. Customers of the company can start industrial modeling the very next day after the conclusion of the contract. For example, MegaFon analysts began work the day after the conclusion of the contract, and a day later there were about two million industrial requests on the platform.

The oneFactor machine learning platform is provided as PaaS (Platform as a Service), all infrastructure resources are included in the license price, which significantly saves time and money on the purchase and deployment of infrastructure. The availability of ready-made tools significantly reduces the start time with the platform.

The platform also has the necessary tools for controlling the quality of these models, characteristics and sources. At each stage of data delivery and creation of new system capabilities, algorithms are used, which, if anomalies are detected, signal "alert" to the monitoring system on the client side. Data quality is monitored both within oneFactor and in MegaFon, Skyeng and Avito monitoring systems.

According to representatives of oneFactor, simplicity, integration modeling capabilities, high performance allow platform users to abandon such foreign software ones as, for example,, SAS Enterprise Miner SAS Real-time Decision Manager SPSS IBM Modeller and. Experian Powercurve