01
Business needs
A Fortune 100 payment card company wanted to revolutionize its customer engagement and market reach through a real-time analytics solution. They wanted to leverage customers’ credit card data for market expansion and utilize an advanced solution that could:
- Offer personalized recommendations to customers in real-time by leveraging machine learning models and predictive analytics
- Develop a self-evolving continuous learning model to detect and adapt to customer preferences over time, ensuring recommendations remain relevant and aligned with shifting trends
- Help manage multiple marketing campaigns and seamlessly distribute marketing offers across diverse channels, both in real-time and through batch processing.
- Elevate customer interactions by extending personalized offers via IVR-integrated service calls
- Provide data-backed insights to the corporate program administrators for relevant and targeted offers, resulting in improved conversion and stronger customer relationships
Ensured 99.995% availability of the enterprise eligibility platform, serving 160+ internal and external vendors, 25+ global markets, and 11K+ campaigns in a year
02
Solution
The Impetus team strategically partnered with the payment card company to unlock multiple analytical-driven marketing use cases and solutions with the following highlights:
- Improved data extraction, curation, and analytics by leveraging Spark and Spark ML operations on the hosted data lake
- Performed real-time data streaming using Kafka and Spark to ingest data into a data lake
- Leveraged Word2Vec and clustering algorithms to derive customer and merchant features for its prediction model
- Managed distributed prediction model training and hyper-parameter optimizations of CatBoost (a family of gradient-boosted decision trees model) model to create non-merchant offers, helping generate a revenue of $250M
- Delivered real-time predictions in milliseconds by training the model with 500+ customer-based features and leveraging the fast execution time of an Apache Spark model
- Enabled an Apache Spark-based solution to provide offers to selective card owners (like providing real-time email offers for customers travelling across the globe)
- Onboarded new channels with only the provided configurations (and no code change) to deliver real-time offers through APIs
- Converted different campaign metadata into uniform metadata by applying minimal changes when integrating with any new campaign management UI
- Used a Kafka-based messaging system to achieve seamless delivery of messages for analytical purposes and Hive tables to store and process messages
- Built an error-resilient system that can retain its point of failure and resume from the same state, saving overall processing time
Processed up to 14M+ card transactions per month with 600+ offers executed daily for marketing campaigns
03
Impact
The enterprise-wide, analytical-driven solution leveraged cutting-edge technologies and helped the payment card company to:
- Simplify the overall marketing process by reducing time-to-market and provide a timely and smooth campaign delivery
- Enable a global scoring system (mapping experience) to drive multiple marketing use cases by analyzing various customer parameters (like spend analysis, age, gender, etc.)
- Process up to 14M+ card transactions per month with 600+ offers executed daily for marketing campaigns
- Enable the enterprise eligibility platform to achieve 99.995% availability, serving 160+ internal and external vendors, 25+ global markets, and 11K+ campaigns in a year
- Improve credit card spending by recommending highly relevant offers to customers
- Extend the real-time APIs support for up to 300 transactions per second with an average response time of 100 milliseconds, maintaining a 99.99% service availability
- Enhance marketing efforts by supporting up to 3.5M+ recommendation requests per day for various campaigns and providing personalized merchant recommendations across 3 marketing channels in 14 global markets
- Improve batch execution and offer delivery to vendors by 10x, reducing from 15 hours to up to 60 minutes
- Deliver 2M+ spend activations in a year by providing real-time, relevant merchant recommendations
- Provide multiple channel and page-placement support on the organization’s website to attract and serve diverse customers
