LG accelerates AI-powered customer analytics transformation
Databricks and Mellerikat combine to improve customer data analysis efficiency
Reduction in AI model development time
Reduction in operations staffing
Increase in conversion rate
LG Electronics is a leading global company in the electronics and home appliances sector, strengthening their competitive edge through innovative technologies and customer-oriented strategies. The company operates in more than 80 countries, with annual sales of approximately 84 trillion KRW as of 2023. With customer data analytics and AI technologies at the core of their strategy, the company has been empowering data-driven decision-making and driving custom business transformation. In particular, they developed the LG Customer Index to analyze customer behavior data and support personalized marketing and strategic decision-making. However, LG Electronics had previously been limited in their ability to effectively leverage their Customer Index due to a decentralized data management framework and inefficient AI model deployment. To address this, LG Electronics combined their edge AI–based MLOps solution, Mellerikat, with Databricks to create an integrated customer data analytics platform.
Low data utilization efficiency due to decentralized data management
LG Electronics uses customer data insights to plan customized products and services and develop customer-centric management strategies, leading to improved customer understanding and satisfaction. Their data-driven Customer Index, which transforms customer behavior data into actionable insights by training AI models, has become a game changer for personalized marketing and strategic decision-making.
Before implementing the Customer Index, LG Electronics lacked a holistic approach to analyzing customer data, making it difficult to manage and leverage insights effectively. Most of the data analysis activities were often one-offs, so the process of building and deploying analytics models was also conducted in an episodic manner. In addition, decentralized data storage made it difficult for employees from different departments to centrally access and manage different types of data, such as CRM and product usage data stored on different platforms. This made it difficult to manage data quality in a consistent and trustworthy manner, which led to inconsistencies in data formats and schemas that created reliability issues and inhibited accurate analysis. Adding redundant data storages to ensure data compliance resulted in wasted expenses and resources due to inefficient infrastructure operations.
The data request and usage process was also inefficient. Data accessibility and sharing were limited, and complex permissions settings and separate data storage for each team led to duplicate data that increased storage costs. The lack of a centralized data catalog made it difficult to monitor data in real time, which led to delays in AI and ML projects, decreased productivity and significantly reduced efficiency of data utilization across the organization.
“Inefficiency in data utilization was a major cause of wasted resources across the organization, and low productivity and satisfaction of data teams,” Keewon Jeong, Senior Researcher at LG Electronics, said. To address these inefficiencies in customer data management, LG Electronics recognized the need for an integrated data management solution.
Implemented an integrated customer data analytics platform by leveraging Databricks
LG Electronics adopted the Databricks Data Intelligence Platform to optimize their customer data management strategy. They leveraged Databricks to efficiently load and manage large volumes of customer data, and systematically implement DataOps, including workspace, ETL (extract, transform, load), workflow and performance monitoring for ML model development. The company combined their MLOps solution, Mellerikat, to build a customer data analytics platform and deploy trained AI models to edge devices for real-time performance monitoring. By combining these two solutions, LG Electronics has been able to efficiently manage the entire process from development to deployment and operation of AI models, support reliable data-driven decision-making and maximize business impact by enabling various departments to create and utilize the Customer Index.
LG Electronics leveraged Databricks’ unified data management to centralize various data sources and create a foundation for systematic analysis of customer data. With Databricks Unity Catalog, the company streamlined data discovery by efficiently organizing and searching datasets, and enhanced data security and met compliance requirements with granular access rights management and integrated governance capabilities. The dashboard allowed teams to visualize data loads and respond quickly to issues, while the workspace provided an independent development environment, facilitating data and code sharing across teams and increasing collaboration. In particular, collaboration between data science and engineering teams has increased, which has greatly improved AI model performance monitoring and speed of troubleshooting.
With this platform, LG Electronics transformed data management and significantly improved business efficiency. The platform provided the basis for analyzing various customer data, which enabled sophisticated customer profiling and strategic decision-making. For example, customer behavior scores for the online brand shop are used to analyze web behavior data and provide personalized search and recommendations, which will improve customer experience and increase conversion rates. By dividing the customer shopping journey into five stages, from sign-up to purchase completion, the platform analyzes customer characteristics at each stage and uses strategies in real time, such as offering discount coupons, encouraging sign-ups and making personalized product recommendations. Subscription lifecycle scores are used to track customers’ journeys with subscription services while predicting the likelihood of re-subscription and recommending the right products to increase customer retention and strengthen performance in the home appliance subscription business. These scores are provided to agents in the Korea Sales Division and are being put to work.
“Databricks’ unified governance and large-scale data processing capabilities played a key role in reliably collecting, processing and managing vast amounts of customer data,” Keewon said. “This helped us to ensure successful DataOps transformation by building a customer data analytics platform that combines our own MLOps platform with Databricks.”
Shortens AI model development time with AI platform created by combining Databricks and Mellerikat
LG Electronics has successfully built a customer data analytics platform by combining Databricks with their MLOps solution, Mellerikat. By integrating Databricks’ robust data management and analytics capabilities with Mellerikat’s edge AI model deployment capabilities, the platform seamlessly handles the entire process of AI model development, deployment and operations. This has accelerated AI-driven customer business transformation and delivered significant gains in data management, AI model development and operational efficiency.
The combination of Databricks and Mellerikat has dramatically shortened the AI model development process. Databricks’ integrated data management capabilities with Delta Lake’s data versioning accelerate the processing of data required for model development. These collaboration capabilities also facilitate communication between data analysts and AI engineers, helping to reduce the time from data request to model development from nine weeks to four weeks. Mellerikat further streamlines model development and operations by leveraging AI content and categorizing customer data challenges into classification, clustering and recommendation system issues. This helped to reduce the process from data request to model development and deployment from four weeks to two weeks, making AI projects more efficient.
Mellerikat’s AI model deployment and operation capabilities have shaped the operating environment more efficiently by connecting with Databricks. They leveraged Databricks workflows to automatically process operational data and connected their learning and inference pipelines with Mellerikat’s edge component to minimize manual intervention. This reduced the manpower required to manage AI models by 40% on a man-month basis, freeing up operations staff to focus on higher-value tasks. In addition, data accuracy and reliability have improved, making model performance management and real-time monitoring more efficient.
With these improvements, LG Electronics’ customer data analytics platform continues to drive performance and efficiency. For example, a campaign that sent out KakaoTalk messages predicting battery replacement time for the A9 vacuum cleaner saw a 3.6x increase in click-through rate and a 1.5x increase in conversion rate. The platform also analyzes customer behavior data to predict subscription lifecycle scores and premium product preferences, and provides personalized recommendations. “We are preparing to expand our AI services based on these achievements and plan to pursue higher efficiency and performance through a new approach based on LLMOps,” Keewon said. “We look forward to continuing our collaboration with Databricks to bring innovative technologies to our business and improve customer experience and business value.”