Meesho, an ecommerce company in India, has developed a generalized feed ranker (GFR) using AWS machine learning services to personalize product recommendations for users. The GFR considers browsing patterns, interests, and other factors to optimize the user experience. Meesho used Amazon EMR with Apache Spark for model training and SageMaker for model deployment. The implementation of the GFR has led to improved user engagement and conversion rates. With AWS services, Meesho’s ML lifecycle runtime has significantly reduced, resulting in increased efficiency.
How Meesho Used Amazon SageMaker to Build a Generalized Feed Ranker
Meesho, India’s fastest growing ecommerce company, wanted to improve the user experience on their platform by offering personalized product recommendations. They used AWS machine learning services, specifically Amazon SageMaker, to develop a powerful generalized feed ranker (GFR) that considers individual preferences and historical behavior to effectively display products in each user’s feed.
Solution Overview
To personalize users’ feeds, Meesho analyzed extensive historical data and extracted insights into browsing patterns and interests. These insights were used to construct ranking models that consider various factors like geography, prior shopping patterns, and acquisition channels. The GFR also takes into account user affinity towards specific items or item properties like price, rating, or discount.
The GFR sources candidate products from multiple channels, including known user preferences, novel and potentially interesting products, trending items, and latest additions. This ensures diverse and relevant recommendations for each user.
The GFR architecture consists of two components: model training and model deployment.
Model Training
Meesho used Amazon EMR with Apache Spark to process large amounts of data and run distributed training at scale. They used Dask, a distributed data science computing framework, to scale out the training jobs across the cluster. This reduced training time from days to hours and allowed for efficient and cost-effective scheduling of Spark jobs.
Meesho used an offline feature store to maintain a historical record of all feature values used for model training. Model artifacts from training were stored in Amazon S3 for convenient access and version management.
They also implemented a time sampling strategy to create training, validation, and test datasets. Various metrics were tracked to evaluate the performance of the model, including area under the ROC curve and area under the precision recall curve.
Model Deployment
Meesho used SageMaker inference endpoints with auto scaling enabled to deploy the trained model. SageMaker offered ease of deployment with support for various ML frameworks and low latency serving of models.
They built an in-house A/B testing platform to monitor A/B metrics and make data-driven decisions. Multiple production variants were deployed on an endpoint using SageMaker’s A/B testing feature. This approach resulted in an approximate 3.5% enhancement in the platform’s conversion rate and increased app open frequency.
Various drifts, such as feature drift and prior drift, were monitored multiple times a day after model deployment to prevent performance deterioration.
AWS Lambda was used to set up automations and triggers for model retraining, endpoint updates, and monitoring processes.
Conclusion
Meesho’s implementation of a generalized feed ranker using SageMaker resulted in highly personalized product recommendations based on user preferences and historical behavior. This improved user engagement and conversion rates, contributing to overall business growth. Utilizing AWS services reduced the ML lifecycle runtime significantly, leading to increased efficiency and productivity for the team.
With this advanced feed ranker, Meesho continues to deliver tailored shopping experiences and fulfill its mission to democratize ecommerce for everyone.
How AI Can Redefine Your Way of Work
If you want to evolve your company with AI and stay competitive, here are some practical steps to consider:
Identify Automation Opportunities
Locate key customer interaction points that can benefit from AI. This could include customer support, sales, marketing, or other areas where automation can improve efficiency and effectiveness.
Define KPIs
Ensure your AI endeavors have measurable impacts on business outcomes. Define key performance indicators (KPIs) that align with your business goals and track them to evaluate the success of your AI initiatives.
Select an AI Solution
Choose AI tools that align with your needs and provide customization options. Look for solutions that can be tailored to your specific requirements and integrate well with your existing systems and processes.
Implement Gradually
Start with a pilot project to gather data and test the effectiveness of AI in your organization. Gradually expand the usage of AI, learning from the pilot and making adjustments as needed.
For AI KPI management advice and continuous insights into leveraging AI, you can connect with us at hello@itinai.com or follow us on Telegram t.me/itinainews or Twitter @itinaicom.
Spotlight on a Practical AI Solution: AI Sales Bot
Consider using the AI Sales Bot from itinai.com/aisalesbot to automate customer engagement and manage interactions across all stages of the customer journey. This AI solution can redefine your sales processes and customer engagement, providing a seamless and personalized experience for your customers.
Explore the solutions available at itinai.com to discover how AI can redefine your way of work.
