User Segmentation Understanding 28 Million MAU, TUES

Toss utilizes TUES(Toss User Engagement Segment) to understand user service usage patterns from a platform perspective. Moving away from existing analysis at the individual service level, it groups users based on which services they prefer and how deeply they use them.

TUES V1 applied the K-Means Clustering algorithm based on each service's usage probability at app open to classify users. Through this, it established strategic criteria for converting simple visiting users into highly engaged users.

TUES V2 addressed the existing limitations and introduced the following changes:

• Changed measurement method: Uses 'usage count per service' instead of usage probability to reflect the depth of usage
• Applied Soft Clustering: Implemented complexity allowing users to belong to multiple segments through the NMF(Non-negative Matrix Factorization) algorithm
• Hierarchical structure: A 3-stage structure flowing from 'service-level engagement → overall app engagement → primary service used', enabling more detailed analysis and Next Action establishment

This segmentation is utilized as a key tool for company-wide decision-making, including improving retention, cross-activation, establishing product growth strategy, and increasing targeted marketing efficiency.