Machine Learning for Ticketmaster Recommendations

Role: At Ticktmaster I led the end to end mobile applications for iOS and Android on the E-Commerce platform for 11M+ users globally.

Problem: One of the core issues that our homepage event feed surfaced on the mobile applications was either repeated events or events that did not have specific relevancy to the user interests leading to consistent scrolling and a reduced Click Through Rate on the homepage, the core focus area for users. The frustration came from user reviews, customer service, and from the quantitative insights we focused on from the behavioral analytics. Users were dropping from the funnel, losing interest scrolling ultimately impacting our conversion rates and return visits.

The Solution: I proactively partnered with the Data Science team to solution for a better output of recommended events for the user in order to garner greater traction. Our goals were to optimize CTR as a secondary metric and conversion and ARPU as our primary metrics. We decided to use data attributes based on weighted values from purchase history, number of purchases, artists, sports events, genre of events, browsing history, and social affinities in order to surface event recommendations for the user which would be in line with their preferences to bring greater engagement and interest to the events that would be surfaced.

The Output: After iterating on the algorithm and testing with beta users, we A/B tested the rollout and started at 5% scale to understand how users were interacting with the new recommendations. I measured the KPI performance between the control and the variant. After seeing an uptick in click-through-rate and conversion over time, we increased the scale of the experiment to 10% and continued to see increased traction. I worked with Data Science to ensure that our KPI's and goals were being met and the outputs were consistently moderated and modified based on what we were seeing.

The Result: After continuing to see positive results over time, we increased the scale of the experiment to 100% and continued the rollout across iOS and Android. We continued to review user scrolling interactions to deeply understand how the event relevancy in the feed reduced scrolling length, scrolling time, and increased deeper interactions and engagement from finding relevant events faster.

Measuring Success: We were able to improve our overall conversion rate by 1% by offering increased event types to our user base and a deeper level of engagement.