Aeroflot implemented an automated system of forecasting of cargo mail tanks of passenger runs

2020: Project completion on creation of AS of forecasting of cargo mail tanks of passenger runs

On February 18, 2020 it became known that the Innodata company completed the project on creation of an automated system of forecasting of cargo mail tanks of passenger runs for PJSC Aeroflot.

In July, 2019 in PJSC Aeroflot the task to optimize commercial loading of runs was set. For its solution it was necessary to optimize completeness of use of a cargo mail quota of passenger runs and to increase the forecasting accuracy of tanks of loading.

Experts of Innodata company who were involved on project implementation analyzed an array of historical data on runs and large volume of unstructured data on loading of aircrafts. Then metrics on which efficiency evaluation of forecasts is made were defined. Results of analytics allowed specialists to develop a technique and criteria for evaluation of tests, including quality evaluation of forecasts. Process of a research happened using machine learning technologies based on the software of Big Data (Hive, Spark ML).

For forecasting of cargo mail tanks of passenger runs often it is necessary to use information on runs on which there are no historical data for training of forecasting models. However the project team designed a system core in such a way that she is capable to process not only such information unrecorded earlier but also to consider data on run loading, the number of passengers, fuel level. Besides, a system reacts to changes in the schedule, replacement of types of aircrafts and other important factors, comments Alexander Sergienko, the chief executive of Innodata company

Except work with Big unstructured Data, project implementation was complicated by need to decrypt special CPM-and LDM- of the telegram which form by preparation of the vessel for a departure. These are data on loading of the airplane along all route and about loading of cargo compartments. As they are formed manually, often contain errors and can have several incoincident versions.

Developers consolidated historical data, constructed the self-trained forecasting models, created the functionality capable to configure and optimize system operation algorithms, and, as a result, provided Aeroflot with the tool for exact forecasting of a cargo mail quota.

As a result of implementation of an automated system of forecasting of cargo mail tanks of the passenger runs Cargo Air in Aeroflot the accuracy of forecast values of loading of aircrafts in 6 months increased by 20%, and the forecasting accuracy of an available free cargo quota grew to 90%. It allowed us to optimize significantly commercial loading of runs in general in scales of all Group, speaks Kirill Bogdanov, the deputy CEO for information technologies of PJSC Aeroflot

2019

End of system implementation of optimization of creation, carrying out and analysis of marketing campaigns

On February 11, 2020 it became known that the ompaniya of Innodat, the Russian developer and integrator of high-tech solutions in the field of customer relations, completed system implementation of optimization, creation, carrying out and analysis of marketing campaigns in PJSC Aeroflot. As the result, a system has unprecedented scale, and is capable to optimize more than 200 campaigns a month, within each of which more than 33 million communications is performed.

The task to provide to Aeroflot the tool for optimization of marketing campaigns due to automation of application of business rules was set for a project team.

For us it is strategically important project which purpose was to provide to marketing specialists of Aeroflot the uniform effective tool for optimization of communications with clients on the basis of approach data-driven. The analysis of last activities allowed to create such tool. It not only reduces operating costs of start of marketing activities, but also provides clients with the optimal personalized offers on available communication channels. Such approach not only positively affects consumer loyalty, but also increases a response to marketing campaigns, commented Kirill Bogdanov, the deputy CEO for information technologies of PJSC Aeroflot

The implemented System based on a single platform optimizes omnichannel approach to communications with clients of Department of marketing of Aeroflot company. A system allows marketing specialists to formulate business hypotheses without involvement of IT department, to check them, to start activities and to create detailed reports with an opportunity to select performance indicators on each campaign. For February, 2020 10 jobs are automated.

The project team worked using management systems of databases of Hive and Spark ML libraries, with a complex of the products IPC Campaign and technology of data exchange IPC Contact Optimization.

The main complexity consisted in study of analytical models because Aeroflot uses numerous data sources in large volumes. Also communication channels are various. Nevertheless, we managed to provide to the customer the tool for achievement of marketing tasks developed on the basis of deep analytics of data and machine learning technologies. It should be noted that a system is easily scaled. So, connection of the next communication channel will allow specialists of Aeroflot to send without additional completions to the client the offer on these communication channels taking into account uniform contact policy, explains Alexander Sergienko, the chief executive of Innodata company

The project started in December, 2018 and came to the end in December, 2019. As a result of system implementation of optimization of creation, sending and the analysis of the marketing companies growth of a response of clients to communications up to 25% and reduction of formal replies from marketing communications of Aeroflot for 15% is expected.

The single platform of marketing communications will act on all territory of the Russian Federation.

Project implementation "Data mart of online sales"

On July 26, 2019 Innodata company, the Russian developer of solutions in the field of analytics on the basis of Big Data, "The data mart of online sales" ("End-to-end analytics") in PJSC Aeroflot announced project implementation. A system allows to analyze efficiency of sales via online channels due to consolidation from different data sources in uniform storage.

As explained TAdviser in Innodata company, the solution is implemented on the basis of the innovation ETL/ BI- products from the largest vendors. At its development were used a system business analysis, visualization and data modeling and structures DB.

Key project task was creation of a uniform data model, sources received from diverse inherently:

• these visits of the website and mobile application. The source is Google Analytics;
• data of sales of tickets. A source – a system of accounting of income;
• data of armorings. A source – a booking system.

In a work progress the analysis of systems, sales processes was carried out and rules of consolidation of the tickets given for digitization of end-to-end sales process are developed.

The reports on key indicators allowing to perform regular monitoring of processes, to estimate results of A/B-testings, to analyze a funnel of sales on the website and in mobile application were configured. Users can create own funnels, compare them among themselves that allows to estimate economic effect of completions of online platforms.

A system allows to prednastraivat the reporting due to implementation of a uniform data model. Further it will give the chance to each user independently to configure the necessary reports and cuts of data. The similar solution will lower load of developers and will facilitate access for users to various data from several systems at once.

Additional benefit of implementation of this project is optimization of the current algorithms of measure calculation due to deeper understanding of processes of the company which it was received during the analysis of systems and processes.

The system of end-to-end analytics – the effective tool for preparation of the uniform reporting for all online sales channels. The automated algorithms of collecting and verification of data will allow to reduce terms of preparation of the reporting and to reduce labor costs, noted Alexander Sergienko, the chief executive of Innodata company