Case Study: GS SHOP boosts ordering customers with Twelve Labs video understanding

GS Shop boosts total ordering customers 57.5% with Twelve Labs

GS Shop, a leading South Korean omnichannel commerce platform, faced the challenge that its video recommendation system could not interpret the content of its "Short Pick" product highlight clips. Their existing engine treated video as a passive thumbnail, blind to the visual context that could reveal customer intent. Manually finding specific moments in hour-long broadcasts was also a time-consuming process for producers. To address this, they turned to Twelve Labs and initiated a proof-of-concept using its Marengo video understanding model via Amazon Bedrock.

The solution from Twelve Labs involved layering a new video-context signal onto GS Shop's existing recommendation engine. The system analyzes videos across four contextual axes—Functionality, Style & Sentiment, Use Situation, and Practicality—to understand what appeals to viewers. This allowed for recommendations based on the scene's context rather than just product tags. The results were significant, with Twelve Labs helping drive a 57.5% increase in ordering customers, a 29.4% lift in conversion rate, and a reduction in broadcast search time from hours to seconds.

GS Shop