Halyk Bank (JSC Peoples' Savings Bank of Kazakhstan) Data Factory
- Customer
- Halyk Bank
- Project manager on the customer side
-
- IT Provider
- DIS Group
- Year of project completion
- 2021
- Project timeline
- January, 2021 — November, 2021
- Project scope
- 184560 man-hours
- Goals
- 1. The strategic goal is to increase the Bank's fee and commission income by personalizing product offers and stimulating transactional activity. 2. To create a single space and analytics and AI toolkit as part of the Data Governance concept on the Bank-wide level. 3. To create a flexible and scalable data platform for launching and testing business hypotheses regarding the Bank's clients. 4. To broadly introduce AI tools for a better customer experience: implementation of anti-fraud systems, AML, an intelligent contact center and other initiatives.
The uniqueness of the project
The Big-Data-based project has shown clear financial results, which ensured business trust and helped the Bank to increase its interest income as soon as three months after the launch of the first client communications based on the data factory.
During the project, a new data platform has been put in place, Data Governance and Data Quality processes have been reinvented, roles have been assigned, and all the Bank’s key corporate sources and data domains have been covered. The approach to collecting corporate data has been completely transformed, business hypotheses have been developed, artificial intelligence (AI) models have been built to increase transactional activity, to improve and personalize customer experience, as well as to reduce customer churn.
We have achieved a full integration of the business in AI tasks along with the use of Big Data tools throughout the entire Bank-Client interaction cycle. The project participants managed to achieve synergy between technological and business tasks.
- Used software
-
The data factory is built around a heterogeneous data warehouse based on Arenadata Hadoop and Oracle DBMS. It is integrated with 14 internal and 20 external data sources using Informatica Data Engineering Integration (DEI), Power Center and Apache Kafka. Data quality is sustained using Informatica Data Quality. All the estimate indicators implemented in the warehouse are described in Informatica Axon, which uses the Informatica Enterprise Data Catalog (EDC) toolkit to track data lineage from 14 internal sources along with their metadata.
A special layer has been built around the warehouse, which uses the data change tracking modules available in the EDC to download real-time data on all card transactions and accrued rewards in the customer loyalty program. Clickhouse, Scylla DB and Elastic Search are used to process this data. A private cloud has been created to deploy decision support systems, with HashiCorp, Vault and Open Stack technologies and Kubernetes and NOMAD container orchestration systems above them.
Contract policy and customer communications are managed using HCL's Unica Journey and Optimize solutions. Real-time decision-making is supported by IBM Streams. Business user access to warehouse data is provided through Qlick BI. For decision-making, real-time data is downloaded from Camunda BPM, with data changes tracking provided through the Debezium open-source toolkit.
- Difficulty of implementation
-
Data Factory has been the Bank's most complex infrastructure project over the past decade. We have completely revised the methods of corporate data sources integration and built Data Governance processes. The data warehouse infrastructure has been fully updated, considering the previous warehouse was implemented as many as 12 years ago.
We have also revised the cross-sales processes with clients in all the Bank’s business areas, both in retail and SME. Lead generation, clients communication and contract policy have been automated
- Project Description
-
A total of 12 teams consisting of 90 employees worked together during the project. They built a heterogeneous data warehouse, introduced 15 data collection and processing components (ETL), developed and implemented 3 key business initiatives, set up 4 communication channels, automated Data Science processes, developed AI models, automated executive reporting, connected 20 external channels to enhance existing information, and assigned 28 data stewards to 60 data domains.
Apart from that, a real-time decision-making system was put in place. A business glossary is maintained to cover all the estimate indicators of the new warehouse. The Data Factory collects data from both internal and external sources, and decisions concerning various actions are made on their basis. Some decisions are made in real-time, including the creation of the client's analytical profile. For other decisions, customer communications data is downloaded from the predictive model daily, including cross-selling, upselling, customer retention, remotivation, etc. This data complements personal interaction histories in mobile banking applications: they are used to create personalized pop-up windows, personalized push notifications, and rewards that are accrued following an analysis of the client’s behavior characteristics and preferences – for this purpose, machine learning algorithms are used. The main principle to be implemented is gentle, unintrusive customer interaction, which is at the same time a win-win for the customers and the Bank.
The project covers both retail businesses, but also corporate businesses and SME.
The project automated customer interaction channels, with both individuals and legal entities. Moreover, ATMs were introduced as another powerful customer interaction channel.
- Project geography
- Kazakhstan
June 2024 was a very dynamic month for AI and data analytics market, marked by significant events, product launches, and industry insights.
On June 4, 2024, the professional IT community Global CIO held its annual meeting with IT leaders from Kazakhstan. IT executives from leading companies in Kazakhstan were invited to the online meeting.
The big news of the past month was the launch of GPT-4o. This new version of generative AI now takes any combination of text, audio, images and video as input, and generates any combination of text, audio and images.
Online VS offline: 5 differences between online and offline retail in inventory management processes
Inventory management plays an important role in retail development as it affects customer satisfaction, competitiveness and overall business performance. Azamat Nirov, Inventory Management product director at Napoleon IT, talks about the main differences in online and offline retail, as well as the factors that determine the specifics of these approaches.