“Chatting during broadcasts will encourage product purchasing” is an important hypothesis for the UI and UX approach for LIVEBUY, LINE’s live commerce service. However, there is a selection bias in which “users who chat are already highly motivated to make a purchase”. Another issue was that there was insufficient data for model generation since the service had just started, otherwise known as a “cold start”.
Using the y-features product, which characterizes and quantifies a cross-section of the LINE service, we were able to augment the lacking data and generate a trend score model with statistical backing. This allowed us to verify that chatting had a positive effect on purchases while excluding selection bias. We’re now planning to create content which will promote chatting from the service side. We’ll take a closer look at our progress towards this goal.