Betting On Big Data: Q&A With Sber’s Alexander Vedyakhin

Alexander Vedyakhin, first deputy chairman of the executive board at Sberbank, shares the Russian bank’s strategy on using data to deliver personalized services.

Global Finance: How do SberProfile and its underlying technology help you create personalized customer experience?

Alexander Vedyakhin: Sber clients are people of different age groups, professions, fields and interests. Our mission is to understand the client’s next need and offer the best solution to this need in the form of customized products and services of companies and partners to our ecosystem. We need to have clear understanding of the customer profile to build a strong personalization system, because nothing will come of it without data and knowledge about customer behavior.

Collecting data from all corners of the Sber ecosystem (not just the bank), being worthy of the trust our customers place in us and taking care of their data is not an easy task. To get a unified customer profile, you need an intelligent, high-performance data platform that consolidates all customer information. What do I mean by “information”? It goes beyond the basic data about a person: their social and demographic profile (place of birth, age, gender, etc.).

The ecosystem positions us to understand exactly how customers use a particular service. For example, we see how much time our online cinema visitors spend watching content per day; the genres they prefer; their favorite directors and films. As for the quality of knowledge, Sber has a major advantage: We see the structure of consumer behavior and can predict customer interests.

Sber combines unique information in each client’s profile, allowing for complex recommendations. For example, we see that you are using Okko, SberMarket deliveries; but you do not have a SberPrime subscription, which helps clients save much money. This means you should be offered one, and this offer will be relevant.

The bank works to improve personal recommendations continuously, focusing on building long-term relationships with the client. It is more important for us not to sell a product to the client right here and now, but to make them want to come back to us. Doing this, we evaluate how our actions today will affect the relationship between the client and the bank in a month, a year or more. Naturally, AI [artificial intelligence] is being used to improve this.

However, AI alone is not enough. You need to work on a portfolio of both financial and nonfinancial products continuously—and this highlights the upside of Sber as an ecosystem that can help a client address a wide variety of life situations. However, working with the ecosystem’s offerings creates new challenges for personalization systems: Approaches that worked well for 200 fintech offerings cannot be scaled up for catalogs with millions of ecosystem products. Therefore we actively experiment with new approaches for building recommendation systems, trying to increase customer engagement in the Sber ecosystem.

GF: What are your top priorities in data management and data control?

Vedyakhin: Our priorities when working with data are those outlined in the goals of our Strategy 2023 and our business goals: We develop the ecosystem, reduce costs and ensure the security of our customers’ data without compromise.

It is clear that Sber is already more than a bank. We are building an ecosystem that provides customers with a seamless, personalized and secure service on a scalable IT platform in any life situation. End-to-end and omnichannel customer journeys in the ecosystem unite the bank and our B2C [business-to-consumer] partners—we need each other’s data to ensure the operation of common services, personalized in view of customer lifetime value, and launch new business initiatives with a short time to market.

Services of this kind need and generate a huge amount of data; and you need to be skilled in handling it, managing it and ensuring its security.

In addition, we are actively building a cloud infrastructure with a short time to data, which will support the joint work of the bank and all ecosystem partners. We anticipate that the ecosystem will be generating millions of transactions per second by 2024. We are already updating the big data platform that will enable the processing of one million transactions per second in 2021 and up to 10 million transactions per second in 2023. In order to ensure that the platform develops in the right direction for the business, we introduced the concept of the business’ co-responsibility for the development and implementation of the data platform and processes. The Retail Unit is responsible for data acquisition and data distribution.

The goals of Strategy 2023 cannot be achieved based on the existing legacy architecture, so we are actively migrating all our systems and services to the target architecture. The same applies to data operations: We have launched an ambitious initiative to structure vast amounts of accumulated data into a universal semantic data layer (USL).

Having a USL will allow teams to focus on designing business products using data and AI, rather than wasting time on nonbusiness issues such as finding the necessary data and consolidating disaggregated data, as well as ensuring quality, completeness, relevance, consistency and availability—and the most difficult aspect: continuous adaptation to change.

This is one of the most ambitious and complex initiatives on the Russian market in the field of building data centers, both technically and organizationally. It will be an interesting experience to share with the world.

We adopted an unconventional approach: We distributed the task among the divisions that own business services generating data and gave them end-to-end responsibility for the data and creation of the USL in their subject areas. We created dedicated agile teams. Data has become an integral part of business products and services. At the same time, the owner has all the expertise and resources necessary to integrate them into the USL. In development, we use data-mesh, model-driven development, and data-vault approaches and principles adapted for Sber.

I would like to touch on the topic of cost optimization. Sber currently has over 244 petabytes of data—the equivalent of 1.6 billion photos uploaded to Facebook. Every second, 200,000 events take place—in terms of volume, that is around 1,500 photos posted on Instagram.

We have set an ambitious goal to reduce the cost of ownership per terabyte of data by a factor of four by the end of 2024. In order to achieve this, we are implementing an information lifecycle management system, gradually transitioning to a custom technology stack for data operations, based on open-source solutions. We are removing vendor lock-in.

Data is one of our key assets, and data security is our competitive advantage as an organization. We value the trust of our clients, care about their data and use the data only with the explicit consent of the client.

To manage and control data operations, we implement data governance processes and are continuously improving the cybersecurity of both Sber’s and our ecosystem partners’ data processing platforms.

For example, this year the board approved the Ecosystem Data Exchange policy, which postulates the key principles of data governance, legal, security and commercial relations in terms of data circulation at the ecosystem level. We closely observe experiences and trends in the regulation of ecosystem data circulation by domestic and foreign regulators.

The chief data officer [CDO] is responsible for introducing a data culture in each Sber business unit. We recently launched the CDO Ecosystem community to share experiences in implementing data management practices and platforms. Data access is strictly regulated. At the same time, during the pandemic, we were able to quickly deploy remote workstations for analysts and data researchers, considering all the relevant cybersecurity requirements.

GF: Please tell us about your conversion of branch offices to last-mile e-commerce delivery centers.

Vedyakhin: Sber offices have SberLogistics order processing and collection points and parcel machines. You can use the service to mail parcels to your friends and family or receive orders from online stores and marketplaces.

The main advantage of SberLogistics is that the collection points and parcel machines are located at Sber branches. People are familiar with these locations—almost everyone knows where the nearest Sber branch is. Many parcel machines are installed at 24-hour service areas by the ATMs. This allows you to plan your time and resources more efficiently when picking up parcels and purchases.

SberLogistics’ own network counts over 5,400 collection points and over 9,000 parcel machines at Sber branches all over Russia. SberLogistics is a Sber ecosystem company offering a full range of logistics services. The company delivers parcels for individual clients, orders from online stores and marketplaces, and documents by couriers, using pickup points and parcel lockers in Sber branches throughout Russia.