This past summer, I worked as a Data Science Intern for the Business Analytics department for the Phoenix Suns basketball team.
Working at the Suns was an amazing experience! I was treated like I was a full member of the team and assigned tasks that seemed to be extremely beneficial for the future of my career. I was exposed to so many different departments and opportunities. While I felt like I would be lonely when I initially decided to move to Phoenix, that was not true at all! Everyone in the organization was so friendly and colloborative and they created a very people-focused environment!
With the Suns, I was part of a group tasked with creating various statistical models, metrics and tools to evaluate ticket revenues, food and merchandise sales and stadium attendances and was part of a team that raised year-over-year revenues despite having an on-court performance worthy of the #1 overall draft pick.
Specifically, one of my summer projects involved creating an interactive, analytical, executive dashboard examining the Suns’ secondary ticket market that team officials, such as the Chief Financial Officer and Director of Revenue, could use to make decisions based on their resell data. Using tools like SAS, SQL and TIBCO Spotfire, I was able to create an interactive arena map depicting seat-level resell trends, as well as provide information on the average resell price relative to various stadium locations, price codes, pricing tiers and across various ticket channels.
Because of this dashboard and my recommendations, the team was able to discover 46,000 previously unidentified illegal seat resales and save over $3.4 million dollars. They also used this dashboard to set pricing strategy for the 2018-19 season and raise ticket prices on high-interest games.
Additionally, through one of my other projects, I used a Breiman-Cutler random forest algorithm to train, test and tune a model that predicted the Suns’ single game sales revenue for every 2018-19 event. The model was used to drive our recommendations for individual-game pricing tiers and allowed the team to dynamically assign game-by-game ticket pricing, which has so far led to another year-over-year increase in revenues. Using previous season data as our testing set and cross validating based on over six seasons of game data, the model had a very low RMSE.
This variable pricing model took in 87 variables, but ultimately used just 11, that ranged from general information to highly specific first-party data, such as date, time of game, and opponent, as well as number of shared Twitter followers between the teams playing, Phoenix Google Trends data, player popularity and NBA merchandise rankings, and projected team win totals.
These projects were really enlightening and taught me skills that will be really valuable for my future career.