Optimizing store operations with the customer in mind

Siloed data sources across industry verticals

Having operated for 150 years and served millions of customers, JKH has amassed enormous amounts of valuable data across multiple verticals.

OCTAVE was formed to help the company’s seven industry verticals derive maximum value from their data, with the ultimate goal of better serving their customers. However, they were not able to effectively leverage their wide variety and large quantum of data due to siloed systems in their business.

JKH, which uses the Microsoft suite of solutions, needed an Azure-native platform that would help unlock the rich insights available to them, and enable better collaboration among their data engineering, science and analytics teams. They circled in on Databricks as the platform that will enable them to centralize and democratize data from various sources — to allow engineers to easily access and build reliable data pipelines for all forms of analytics, scientists to scale their exploration and model training, and analysts to derive meaningful business insights to improve store operations through better inventory management and customer demand forecasting.

Assisting OCTAVE in accelerating this transformation in cloud infrastructure was John Keells Information Technology, a Microsoft Gold Partner, which implemented the platform architecture.

Seamless integration for effective cross-team collaboration

Azure Databricks enabled the OCTAVE team of over 30 data engineers, data scientists, analytics delivery managers and visualization analysts to work on a single platform, managing the full data lifecycle from data preparation all the way to delivery of analytics and insights, introducing new cross-team synergies and standardized practices to their operations.

Chanaka Bandula Hewa, a data engineering expert at OCTAVE, said, “One of Databricks key values is how it has facilitated cross-team collaboration. The ability to easily share workspaces, clusters and jobs through a single interface, and deploy multilingual collaborative notebooks means employees can now securely access the self-service platform to effectively collaborate across teams.”

Being on the Microsoft Azure ecosystem allowed fuss-free integration of services such as MLflow and Delta Lake into the existing tech stack. The OCTAVE team appreciates Databricks’ simple architecture with a single compute platform that supports most of the use case needs and allows integration with wider group services, including Azure security frameworks, reporting tools such as Power BI, version controls and integration with DevOps for end-to-end management capabilities. This allows the data analysts to leverage holistic data insights to create reports and dashboards so they can make informed decisions that improve store operations and boost customer engagement.

The use of MLflow has greatly streamlined the machine learning lifecycle — improving experiment tracking and model project versioning at OCTAVE. Data scientists previously ran individual experiments, tested algorithms and tracked the results on Excel spreadsheets that were highly error prone and time-consuming. With MLflow, teams now use standardized tools, processes and commands on a single repository for recording experiments and model objects. This allows for much improved visibility, which is especially useful when making comparisons across experiments.

Leveraging data to unlock actionable customer insights

OCTAVE is running a roadmap of over 40 strategically significant advanced analytics use cases aimed at solving business problems across the organization.

One such use case was piloted for supply chain optimization at Keells — the supermarket chain within the John Keells Group. One of the biggest challenges that retailers face is anticipating customers’ purchasing needs. Different demographics in each store location mean demand for goods can vary, and deciding how much space to allocate to each category can be tricky.

Traditionally, Keells outlets would allot a common assortment of goods based on the size of the store. The decision of whether to stock a range of products was made based purely on store size — a process that presented a clear opportunity for data to inform and optimize.

Applying the advanced analytics model enabled by Databricks helped Keells to cluster the stores based on features related to customer behavior. The pilot showed material improvement in “dry categories” revenue and a significant reduction in the time taken for the annual category review.

“Databricks is now the primary platform for model development at OCTAVE, and usage is even extending beyond OCTAVE into the business units that are finding it beneficial,” said Yolan Seimon, Vice President, John Keells Group, and Head of Data and Advanced Analytics.

Looking ahead, John Keells will continue to leverage Databricks to enable more use cases through the unification of data engineering, machine learning and data analytics, giving rise to the rapid adoption of advanced analytics designed to improve business operations and the customer experience. “We’re excited about the lakehouse vision that Databricks has set forth,” said Seimon, “and we look forward to continuing to move toward that direction as we democratize data throughout the company.”