CorEnroll Recommender

• Location: Reykjavík, Iceland
• Project Lead: Atul Ranjan
• Project type: Advanced Analytics, Machine Learning, Narrow AI, MLOps
• Project Partner: CloudTech
• Project Client: CorEnroll
• Market: USA and Canada

Background

In the bustling world of e-commerce and product-revenue management, platforms often offer a vast and diverse array of products. With this diversity comes the challenge of helping customers navigate through countless options to find products that meet their needs and preferences. This is where recommendation systems come into play.

Enhancing Customer Experience

Recommendation systems personalize suggestions based on user behavior, making shopping easier and more enjoyable. This increases the likelihood of customers finding desired items, boosting satisfaction and loyalty.

Driving Sales and Revenue

These systems drive sales by boosting cross-selling and up-selling opportunities. For example, after purchasing a smartphone, a customer might be suggested compatible accessories, increasing order value and overall sales.

Improving Product Discoverability

Recommendation systems improve product discoverability by highlighting less visible items for the right customers, ensuring niche products reach interested buyers and optimizing inventory turnover.

Gott Data had the pleasure of working with CloudTech, a leader in custom software development focusing primarily for insurance services in the United States. In this case, the need for recommendation AI agents was conceived for CorEnroll platform which allows brokers and insurance reps across the USA to sell insurance packages to customers in all states. As the insurance packages and offerings have to follow the U.S. State laws and various healthcare requirements, insurance reps spend a bulk of their time searching for the right insurance product to sell to the customer. It is a critical task for insurance reps to have right products at their finger tips to sell, so they can sell them to customers quickly and maximize revenue during the annual month-long enrollment period. On the other hand, when the customer purchases the right product (and at a lower price than market value) that matches their needs, they are more likely to be 'happy and have a peace-of-mind' (and become a returning customer!) and avoid unforeseen costs when they have to switch plans outside of the enrollment period during the enrolled year.

Objective

Gott Data collaborated with CloudTech to achieve several key objectives for CorEnroll. These included enhancing data organization and business intelligence capabilities to streamline operations and improve decision-making processes. By leveraging advanced customer segmentation, Gott Data provided data-driven insights that personalized customer interactions and drove sales.

The partnership also focused on improving customer convenience through sophisticated recommendation systems and optimizing portfolios to maximize returns. Additionally, Gott Data was committed to increasing data awareness, ensuring that all stakeholders were informed and able to harness the power of data effectively.

Scope

The project scope included:

1. Data Integration: Combining data from various sources to create a unified, comprehensive dataset.
2. Data Transformation: Cleaning, structuring, and converting data into a usable format for analysis.
3. Business Intelligence & Insights: Developing tools and dashboards to provide an overview of key metrics and insights, facilitating informed decision-making.
4. Customer Segmentation: Analyzing customer data to create distinct segments for targeted marketing and personalized services.
5. Portfolio Optimization: Enhancing portfolio management strategies to maximize returns and minimize risks.
6. Recommendation System: Implementing advanced recommendation algorithms to improve customer experience and drive sales.
7. MLOps Pipeline: Establishing a robust MLOps (Machine Learning Operations) pipeline for continuous integration, deployment, and monitoring of machine learning models.

Challenges

The project faced several significant challenges:

1. Huge Sums of Data: Managing and processing vast amounts of data required robust infrastructure and efficient data handling techniques to ensure smooth integration and analysis.
2. Skewed Data: The data was largely skewed towards one product, complicating the process of creating balanced and representative models. This required additional efforts in data sampling and adjustment to ensure accurate and fair recommendations.
3. Pushing Manual Plans in the Recommender through Sampling: Integrating manual plans into the recommendation system involved careful sampling to maintain the integrity and relevance of recommendations while incorporating expert input.
4. Fast Inference Time: Achieving fast inference times was critical to ensure that recommendations were delivered promptly, enhancing the user experience and maintaining system efficiency.
5. Data Lake and Data Transformations: Building and maintaining a data lake, along with performing necessary data transformations, posed challenges in terms of data organization, storage, and processing.
6. Low Model Size: Keeping the model size low without compromising on performance was essential to ensure efficient deployment and operation, particularly in environments with limited computational resources.
7. Factor-Based Recommendations: Implementing factor-based recommendation algorithms required sophisticated techniques to accurately model user preferences and product characteristics, balancing complexity and performance.

Solution

Gott Data addressed the project challenges with innovative solutions:

1. Huge Sums of Data: Implemented scalable data processing frameworks and optimized data pipelines to efficiently handle large volumes of data, ensuring timely integration and analysis.
2. Skewed Data: Employed advanced sampling techniques and data balancing strategies to mitigate biases and ensure that recommendations were representative across all product categories.
3. Pushing Manual Plans in the Recommender through Sampling: Developed a hybrid recommendation approach that seamlessly integrated manual plans through sophisticated sampling methods, maintaining the integrity and relevance of recommendations.
4. Fast Inference Time: Optimized model architectures and implemented efficient algorithms to achieve rapid inference times, ensuring that recommendations were delivered promptly to enhance user experience.
5. Data Lake and Data Transformations: Established a well-structured data lake architecture with streamlined data transformations, facilitating seamless data access and processing for enhanced business intelligence and operational efficiency.
6. Low Model Size: Implemented model compression techniques and optimized algorithms to maintain low model sizes without compromising accuracy or performance, enabling efficient deployment in resource-constrained environments.
7. Factor-Based Recommendations: Leveraged advanced factorization techniques and collaborative filtering algorithms to deliver personalized recommendations based on intricate user preferences and product characteristics, enhancing recommendation quality and customer satisfaction.

Delivery

Through these solutions, Gott Data successfully navigated the complexities of the project, ensuring robust performance, scalability, and enhanced business outcomes for CloudTech. While the initial brief of this project was to build one recommendation system for the platform, during our analysis phase using our approach we identified three critical phases where insurance reps made decisions on what products would be best for their customers. Based on these findings, we deployed three recommenders in the platform:

A. Top trending products based on insurance type in their state - these recommendations are based on top sales in US$ and number of sales over the enrollment period by them and other insurance reps matching their profile and customers. This provided them quick access to products that they could offer to potential customers quickly.

B. Advanced Similarity-based recommendations - these recommendations are based on filters applied by the insurance reps which lets them narrow down packages based on their customer needs and their details. Recommendation AI agents find similarities between search parameters and match weights between these parameters and particular products, essentially recommending products that matches most to the needs of the customer.

C. Complementary product recommendations - these recommendations are recommended as an add-on during the checkout process, recommending complementary products allowing the potential for agents to bundle other products into the offer. This significantly increases the cross-selling and up-selling opportunities, and consequently increases revenues for the platform.

Gott Data delivered these recommendation systems through API Endpoints deployed in the AWS Cloud for fast inference along with MLOps Pipeline and complete documentation for the project. Gott Data also delivered Business Analytics Dashboards and data findings from throughout the project, which allows the CloudTech team to monitor the performance of these algorithms & also compare them with actual real sales figures to make sure the recommendations are aligned with company goals. In addition to this, Gott Data organized and held a few knowledge transfer sessions with the incredible CloudTech software team for seamless onboarding. This also included ways to tweak these recommendation AI agents so that the CloudTech team can adapt them for the ever-changing real-world scenarios.

Impact

The project led by Gott Data significantly improved the platform's operations and customer satisfaction. By optimizing data processing, the project enabled faster analysis and more informed decision-making. Advanced techniques ensured unbiased recommendations across products, enhancing user engagement. Integrating manual plans into the recommendation system improved accuracy and customer loyalty. Faster inference times boosted user experience with timely recommendations. These enhancements underscored the project's success in driving tangible benefits for CloudTech, cementing its leadership in the health and insurance sector. In addition to this, the data reports provided by Gott Data proved valuable in providing a deeper understanding of the sector, the user base, and the existing plans through advanced analytics techniques.

Besides these, there have been two significant impacts that guarantee continued competitive advantage for the platform:

1. Increase Revenue and expansion of portfolio offerings as reps are now allowed to sell more, and diverse products that they were previously unaware of.
2. Reduced workload, and higher customer & insurance reps satisfaction

On the other hand, because of our rigorous approach in this project, we have additionally added more value into various critical areas within operations and customer engagement:

1. Enhanced Data Processing: Improved data processing frameworks enable CloudTech to handle large datasets efficiently, supporting faster analysis and decision-making in insurance plan management and customer service.
2. Personalized Recommendations: Advanced recommendation systems provide tailored suggestions for insurance plans and services based on individual user profiles, enhancing customer satisfaction and retention.
3. Operational Efficiency: Streamlined data transformations and robust data lake management ensure seamless access to actionable insights, optimizing operational workflows and resource allocation.
4. Risk Management: Effective portfolio optimization strategies minimize financial risks through data-driven insights and proactive monitoring of insurance claims patterns, ensuring sustainable business growth.
5. Customer Segmentation: Enhanced customer segmentation capabilities allow for targeted marketing campaigns and personalized service offerings, improving engagement and conversion rates.
6. Machine Learning Operations (MLOps): Implementation of MLOps pipelines ensures continuous deployment and monitoring of machine learning models, maintaining high performance and reliability of recommendation systems.
7. Business Intelligence: Comprehensive business intelligence tools provide stakeholders with clear, actionable insights into market trends, customer preferences, and operational performance, driving strategic decision-making and competitive advantage in the health and insurance sectors.

Application

For any organization considering projects to understand their product offerings better, increase competitive advantage over their competitors, and increase their sales through B2B and B2C channels, advanced analytics projects are a solid investment into their future growth. In this project, even before we started writing the code for these customer recommendation AI agents, we had already provided invaluable insights into their platform, transactional trends, and customer buying behavior. With our bespoke recommender AIs, we have solved critical human-focused pain points of their platform and used advanced analytics to provide 'needle in the haystack' insights to decision-makers that give them increased confidence in meeting their organizational goals. Recommendation systems are great projects for any business, even though the e-commerce application is the most prevalent and well-known use case. Our advanced analytics methodology removes the instinct and bias from the process, so it empowers people to make best decisions during their work processes.

If this resonates with your business or organizational needs, drop us a line and we will be happy to help you!

For more information about our advanced analytics services: learn more

For other information: Visit our Website

Follow us on LinkedIn: Gott Data