Title

AWS re:Invent 2023 - Delivering ML-driven campaigns with RudderStack and Amazon Redshift (ANT221)

Summary

• Eric Dodds, head of product marketing for RudderStack, introduces Mike Spurdy, VP of Engineering at StatPearls, who shares their journey from data chaos to ML-driven marketing campaigns.
• StatPearls, a healthcare and technology company, faced challenges due to a complex product structure and a complicated data schema involving numerous medical publications, questions, and professional certifications.
• Initially, the company struggled with organizing data and relied on ad hoc SQL scripts for insights.
• The first step towards improvement was integrating clean first-party data into Amazon Redshift, with RudderStack helping to map data flows and third-party tool integrations.
• A significant challenge was identity resolution within the data warehouse, which was addressed by using RudderStack Profiles to create detailed customer profiles and KPIs.
• With a complete customer view, StatPearls could better understand customer lifetime value and subscription renewal rates, leading to a 4x increase in Google Ads budget based on newfound insights.
• The company then moved from rearward-looking analytics to predictive actions using machine learning models integrated into RudderStack Profiles, starting with churn prediction.
• The churn scores were appended to customer profiles in Redshift and used to inform targeted marketing campaigns through various tools.
• Future plans include incorporating more ML models, such as likelihood-to-buy predictions.
• The investment in data infrastructure with Redshift and RudderStack is justified by the significant impact on sales and marketing efficiency.

Insights

• The transition from content creation to monetization required a shift in focus towards data organization and marketing strategies for StatPearls.
• The integration of RudderStack and Amazon Redshift provided a scalable solution for managing complex data and marketing stacks.
• Clean data collection and identity resolution are critical for effective marketing and customer insights.
• The use of RudderStack Profiles within Amazon Redshift allowed for the creation of dynamic customer profiles, which could be easily updated with new attributes without extensive SQL coding.
• Machine learning models can be effectively integrated into existing data workflows to enhance predictive marketing capabilities.
• The ability to directly connect churn scores and other predictive traits to marketing tools streamlined the process of executing targeted campaigns.
• The case study demonstrates the tangible ROI of investing in a robust data infrastructure, which can drive sales growth and operational efficiency.