Multi-Cohort Behavioral Telemetry and Longitudinal Attrition Metrics Across Gated LMS and Open-Access Streaming Environments (2014–2025)

This data descriptor documents a unified collection of four longitudinal datasets capturing user interaction telemetry, retention dynamics, and attrition boundaries across self-paced digital mind-body interventions (relaxation, mindfulness, physical yoga, and yoga philosophy) recorded between May 2014 and December 2025. Telemetry was extracted from two parallel technological architectures: a gated, structured Learning Management System (LMS, Udemy) and an ungated, algorithmically driven video streaming environment (YouTube). Collectively, the repository comprises 1,787 anonymized individual student progression logs in the LMS environment, aggregated country-level demographic distributions across up to 52 territories, granular video-level retention and sequential playlist drop-off metrics across 107 instructional modules, and multi-year annual view timelines. The data architecture is specifically structured to support survival analysis (e.g., Kaplan-Meier estimation), activation gap modeling, gaming/speeder detection, and cross-platform attrition comparisons. All tabular files strictly enforce Tidy Data principles, include automated schema validation via Frictionless Data specifications (datapackage.json) alongside semantic manifests (ai-manifest.jsonld), and are publicly released at Zenodo under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.