Activate 2018

Type Ahead (TA) is a universal feature that large e-commerce companies use to automatically complete a user’s query based on a prefix. Being the U.S.’s 4th largest e-commerce site; TA service is one of the most important service that delivers optimal user experience.

TA results are often based on previous popularity, conversion rate of the queries. However, such hardcoded rules are not the most optimal. With massive TA usage data being generated each day, we employ a machine learning model to learn the “why” behind what drives users to select the suggested terms from TA. This model learns important features behind each decision ranging from statistical probabilities to semantics and seasonality to past user satisfaction. Our model then re-rank TA suggestions such that only the most relevant results are portrayed to the user, thus creating a seamless customer experience. This talk will showcase the effects of machine learning to create a more powerful and robust online shopping experience.