Aeroflot unrolled the platform on work with customer appeals on the basis of Big Data technologies

2018

On February 8, 2018 the Aeroflot company announced deployment of the platform on work with customer appeals, having used for this purpose algorithms of predictive analytics of data.

In Aeroflot the platform for work with customer appeals is implemented

Customer appeals – a key source of information on quality of service of Aeroflot. Increase in loyalty is impossible in the absence of smoothly running and quick reaction on the comments left by passengers, complaints and queries arriving from different communication channels (including social networks). The loyalty increases if the company is open for notes and offers, and clients are sure that their wishes will be heard and apprehended.

Project Objectives

It was necessary for increase in efficiency, speed and quality of processing of client requests:

• create a single platform on processing of addresses from different communication channels, them

intellectual classification and routing;

• provide the tool allowing to trace online key trends in

social networks, including a feedback about PJSC Aeroflot from leaders of opinions (popular bloggers, journalists and so forth) and if necessary to use all available channels communications for preserving of image of the company.

Project Progress

The Integro Technologies company suggested Aeroflot airline to consider the possibilities of the BigData technology providing consolidation of historical data, creation and regular training of models for intellectual classification of addresses, search of implicit interrelations and doubles.

Having conducted a research of the market of the solutions supporting BigData technology, specialists of Aeroflot came to a conclusion that platforms IBM BigInsights and IBM PureData are the most functional and flexible and meet requirements of airline.

Within the project the fundamental functions of a system allowing to configure and optimize flexibly further algorithms of its work were designed and implemented.

First, process of monitoring of references of Aeroflot on social networks, trackings of their tonality and dynamics of change after different information occasions is implemented. The analysis of responses from social networks was carried out on an array of all Russian-speaking segment since January, 2014. Data retrieveds were used for machine learning that allowed to define in real time category and tonality of specific mentioning, helping responsibles it is correct to define priorities of response to addresses in social media.

Secondly, on the basis of the saved-up historical data the mathematical models allowing when obtaining the address from the client on any communication channel were constructed (with a high share of probability) to automatically determine an address subject and also the division responsible for processing and solution. Such approach allowed to reduce use of resources for the analysis and reassignment of the received requests. Moreover, in the course of work, a system analyzes "fresh" data and samoobuchatsya continuously for increase

Having integrated these two blocks, it was succeeded to implement end-to-end process, since obtaining the address in a free form on one of social networks, and finishing with providing irrefragable answer to the Client, and company management – the detailed analysis of the reasons of the event and the uniform reporting under addresses for acceptance of management decisions.

Project Results

A system adjusts about 5% of wrong subjects of addresses and independently defines a subject in 85% of cases.

Due to rapid response to the most resonant information occasions, the loyalty of the potential and existing clients of Aeroflot significantly increased.

The assessment of a customer loyalty was carried out on selection of addresses of social networks and from among the addresses which arrived through Single window. Processing speed of "standard" addresses increased by 7% already a month later after the beginning of trial operation. Also a system allows to automatically determine tonality and to route more than 40 thousand addresses monthly. More than 130 thousand references gather in social networks, on them is carried out aspect santimentny the analysis for the purpose of determination of the most concerning passengers of questions. Everything together it creates solid base for prevention of emergence of conflict situations and considerable decrease in terms of consideration of addresses, - Kirill Bogdanov, CIO of Aeroflot told.

The Integro Technologies company performed industrial implementation of BigData technology for automation and consolidation of processes on a joint of traditional channels of communication with clients and social services.

According to Kirill Bogdanov, the most important business result — consolidation of complaints from all channels of addresses (e-mail, a single window, representations, offices) and references from social networks in a single platform. It allowed to collect work with addresses in uniform business process, to apply machine learning to automatic analysis of subject of the address for the purpose of the subsequent routing and also to conduct monitoring of responses and addresses on social networks in real time.

Our ultimate goal — to reduce terms of processing of addresses, to communicate proactively with clients on social networks, to improve client experience. The client happy with service will return to us yet not once for this reason we appreciate an opportunity better to study and become at each address, - he noted.

At the same time, project deliverables showed unconditional efficiency of technology that is confirmed by plans of the customer for expansion of its use.

Basis of all current and future future changes when approaching to work with clients are technologies of work with Big Data. Thanks to BigData Aeroflot has new opportunities of conducting effective business. This project can unambiguously be considered break of year. In world aircraft there are already CRM projects, but all of them — without global collecting and data analysis from social networks. And exactly there the behavioural profile of the client accumulates now. Regarding work with clients Aeroflot took a step forward, having increased incredibly large volume of sources of a feedback from clients for receiving the most exact and reliable picture of relationship, timely response to complaints and addresses. All this leads to increase in level of loyalty, and finally to growth of sales volume of the company, - Kirill Bogdanov added.

2016

Until the end of 2016 the Integro Technologies company will have to develop and implement an automated system of complex market research which will issue to carrier of the recommendation in Aeroflot how to increase the cost of rates when preserving volume of armorings.

Follows from the project documentation published in June, 2016 that a system will be able to collect information on the current existence of places and cost of tariff classes at competitors, their availability to purchase via portals aggregators of offers. Data collection should be performed automatically at least two times a day.

On the basis of collected data a system will allow to create predictive models on rates on straight lines and transit international routes and to issue recommendations about change of rates. Model operating time for creation of the forecast for all directions and all classes of service should not exceed two hours.

The new IT system of Aeroflot will analyze proposals of competitors and on the basis of it to issue recommendations where and as it is better to raise rates

The recommendation will include the new cost of a tariff class, validity period of new cost and the list of justifications in business terms. Economic effect of the recommendations offered by a system should exceed effect of use of a simple algorithm of following to the competitor's rates using simple correction factor, it is specified in requirements to a system.

One of solution specifications is existence of means in it information visualization and interactive data analysis by such parameters as the list of changes in value of tariff classes for different carriers, dynamics of changes on carrier, comparison of schedules, dynamics of change of real availability of tariff classes, dynamics of scope changes of armorings of Aeroflot in a binding to changes of rates of other suppliers and other events, the forecast of a change in value of tariff classes for the direction and many other things.

^[1] from suppliers on system development Aeroflot announced request for proposals at the end of May, 2016. The starting price of the contract was 252.3 million rubles.

Won Integro Technologies suggested to perform work for 229 million rubles. Also participated in the tender Technoserv Consulting and "MFM Integration".

In the general project budget value of the rights on the software is 115.5 million rubles, the cost of a hardware platform – 68.7 million rubles, the cost of works on development and deployment of a system – 44.8 million rubles.

In technical requirements to a system it is specified that it should be implemented based on the platform IBM including rights to use of such software as IBM Biginsights for Apache Hadoop, SPSS Modeler and InfoSphere DataStage. IBM BigInsights, in particular, will provide data storage, functional requirements, necessary for implementation, IBM InfoSphere DataStage - will provide data collection, and IBM SPSS Modeller will be used for creation of analytical models.

A system should become a part of a common information space of Aeroflot and provide all necessary information for other information systems used in the company.

Notes