Introduction
This case study presents a business scenario where a client faced challenges in recommending products to customers. The client, an undisclosed company in the Industrial Machinery & Equipment industry, aimed to enhance customer experience, increase sales, and promote multiple product purchases. To address this challenge, the SmartCat team developed and implemented a Frequently Bought Together recommender, leveraging transaction data analysis and machine learning techniques.
About the Client
The client, an undisclosed company in the Industrial Machinery & Equipment industry, specializes in transforming mechanical vending machines and kiosks into smart IoT devices with touchless technology, AI-powered engagement, and crypto payments. They focus on providing and servicing automated machines that sell a variety of products, including snack foods, soft drinks, cigarettes, newspapers, and other merchandise.
Problem
The client faced the issue of manually or randomly recommending products to customers, resulting in low personalization and missed revenue opportunities. The pain points and obstacles included suboptimal customer satisfaction, underutilization of transaction history, and limited cross-selling potential.
Solution
To overcome these challenges, the SmartCat team proposed the implementation of a Frequently Bought Together recommender. This solution involved analyzing the extensive history of transactions to identify patterns and extract insights on frequently purchased product combinations. By implementing this recommender, the client aimed to enhance the user experience, improve sales, and boost customer satisfaction.
The strategies employed by the SmartCat team included:
- Data Analysis: Thoroughly analyzing transaction data to identify patterns, seasonality, and geolocation features.
- Machine Learning: Utilizing Python, Spark, and MLflow to develop and train a recommender model capable of predicting frequently bought products based on the user’s selection.
- Integration: Collaborating with the client’s technology team to seamlessly integrate the recommender into their existing systems.
- A/B Testing: Conducting rigorous testing to evaluate the performance and effectiveness of the recommender in real-world scenarios.
Results:
The implementation of the Frequently Bought Together recommender yielded significant outcomes for the client. The key metrics used to measure the impact of the solution included:
- Total number of bundle transactions: Increased by 7%.
- Sales: Improved by 23%.
- Bundle transaction ratio: Showed a 10% increase.
- Revenue: Recorded an increase of $4700000.
- Average purchase revenue per bundle transaction: Increased by $70.
These quantifiable results demonstrated the effectiveness of the recommender in driving sales, enhancing customer satisfaction, and promoting multiple product purchases.
Smart Tip:
To ensure the validity and relevance of the solution, it is crucial to conduct A/B testing. This will provide concrete evidence of the recommender’s performance in real-world scenarios. Additionally, comprehensive data analysis, including seasonality and geolocation features, can help generate more relevant recommendations.
Smart Fact
According to industry studies, implementing personalized recommendations based on frequently bought together patterns can lead to an average sales increase of 10-30% for e-commerce businesses.
Technologies Used
The SmartCat team utilized various technologies and tools to address the problem and implement the solution. These included Python, Spark, Databricks, MLflow, Azure, and Delta Lake. These technologies enabled efficient data analysis, machine learning model development, and seamless integration with the client’s existing systems.