South KoreaRetail and e-commerce2026 – ongoingGS SHOP
Video-context recommendations and broadcast scene search
TwelveLabs reported 57.5% more ordering customers after GS SHOP added video signals
Problem
GS SHOP's recommendations could not interpret the visual context of shopping clips. Producers also searched broadcast footage manually.
Approach
GS SHOP tested TwelveLabs Marengo through Amazon Bedrock and moved into production. It combined video-context vectors with customer intent and existing product data. Claude removed brand and product names from descriptive signals.
Results
Takeaway
Video signals extended an existing recommendation engine. If video processing was delayed, recommendations fell back to product data.
Categories
- Size
- Large enterprise
- Industry
- Retail and e-commerce
- Business function
- Marketing and salesProduction and quality
- Innovation type
- Process innovation
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
TwelveLabs described production recommendations informed by video context and faster broadcast search. It reported increases in ordering customers, conversion and clicks, without specifying an evaluation period.
Full details are in the original sources.