'iRecommend' wins GfK's NextGen Data Science Hackathon
The idea for a service that would give buyers recommendations for products they cannot live without has won four computer science students in the Ying Wu College of Computing 1st place and a $5,000 grand prize as part of GfK’s annual NextGen Data Science Hackathon Competition.
“iRecommend” is a proposed data-driven system for making FMCG (Fast Moving Consumer Goods) ecommerce experiences more personalized and relies on predictive analytics to recommend products to shoppers based on their purchase histories and individual characteristics.
GfK challenged undergraduate students to develop strategic recommendations for new products or services in the shopper insights space. Competitors had access to data from GfK’s National Shopper Lab (NSL) – which captures loyalty card activity from 96 million US shoppers – and other sources. The teams needed to show excellence in data science and analytics above all, along with a plans for an innovative offering.
Finalist teams from the US and abroad had five minutes each to present their ideas to a panel of expert judges that included leadership and data science experts from Pepsico, Oracle and Ferrero.
Dang Huynh, Khoa Nguyen, Tuan Phan and An Tran, all B.S. in Computer Science students in their junior year, used assigned datasets and demographics of consumer purchases over three years to identify patterns that would provide insights to businesses. The team then built two recommendation engines to match products with each respective user. The first engine would suggest relevant cereal brands while the second identified appropriate categories based on age, gender and ethnicity.
Assistant Professor Pan Xu, the team’s faculty advisor, helped guide their research and develop algorithms to accommodate for certain variables such as how behavior might be influenced by different seasons in the year. (One might wonder if the choice of your favorite sugary cold cereal from childhood over hot oatmeal ever goes out of style regardless of age or temperature.)
Tran, the project leader, noted that the 10-day duration of the competition simulated the timeline of a real-world project and embraced the challenge of presenting enough information at the end to give the judges an impression of the products in the allotted five-minute limit.
“We were beyond pleased with the diversity and high quality of responses to this year’s NextGen challenge,” said Rolfe Swinton, GfK’s Director of Data Assets for North America. “The teams’ knowledge of data science and analytics continues to grow each year, and we were glad to see students from a variety of majors and backgrounds taking part. [These teams] made strong cases for concepts that have real marketplace value and relevance.”
Now in its 11th year, the NextGen Competition gives undergraduates firsthand exposure to solving real-life business problems with consumer insights and data. Four years ago, GfK North America reimagined the competition as a 10-day hackathon, in which students mine raw data sets for relevant insights and then convert them into business guidance. The change reflects a radical transformation in market research, in which data integration and predictive analysis now play dramatically larger roles.