Empowering Decisions: Azure Databricks Transforms CPG Strategy

Client profile

A multinational consumer packaged goods (CPG) firm headquartered in London, UK, boasts a diverse assortment of products ranging from baby food and beauty items to healthcare products and soft drinks.

Business challenges

• The existing recommendations system operates on a traditional SQL platform, generating approximately two crore recommendations per country each month.
• Complex logic and machine learning algorithms drive the recommendation generation process, requiring substantial manual effort (approximately four person-days per country) and time.
• The client aims to improve the process by reducing processing time, eliminating manual intervention, and streamlining assortment planning and decision-making.

LTM solution

• LTM facilitated the migration of the recommendation generation process to Azure Databricks.
• Automated the end-to-end recommendation generation process, utilizing Scala coding for efficient data retrieval, consolidation, and machine learning algorithm implementation.
• Azure Databricks enabled seamless model refreshing with new data, eliminating the tedious process experienced on the SQL platform and ensuring the latest insights are always available.
• Advanced deep learning and network algorithms enhanced the platform's capabilities, enabling more refined analysis and recommendation generation.

. Business benefits

• Real-time recommendation generation utilized 29 days of sales data and provided actionable assortment planning and optimization insights.
• Improved on-shelf availability (OSA) of products led to a significant uplift in revenue as products aligned better with consumer demand.
• Drastic reduction in churn time from four person-days to just one day per country enhanced operational efficiency and agility.
• Increased sales representative productivity enabled them to focus on strategic initiatives rather than manual tasks.
• Reduced error rates and elimination of manual interventions ensured the accuracy and reliability of recommendations.