Diagnosing Samsung's CIC learning platform — a data-driven case for LXP transition.
CIC is Samsung's enterprise-wide learning platform serving employees across multiple subsidiaries. This capstone project combined systematic content analysis with ARCS-based UX diagnosis and LXP transition strategy — proposing data-grounded improvements to close the gap between what CIC offers and what employees actually engage with.
A platform caught between LMS and LXP — with user behavior telling a different story.
Post-pandemic, Samsung employees' learning expectations shifted. Accustomed to YouTube, Netflix, and short-form content platforms, users now expect personalized recommendations, intuitive interfaces, and content that respects their time. CIC had begun incorporating LXP-like features — feeds, campus communities, curated playlists. But the core architecture still reflected a compliance-first LMS.
How do we move from a platform that manages learning to one that drives it — using real content performance data to close the gap between what CIC offers and what employees actually engage with?
200+ contents. Four counterintuitive findings.
CIC platform content was systematically analyzed via HTML source extraction — capturing views, engagement signals, content length, and section classification across 200+ published items.
From platform observation to strategic diagnosis.
The project followed a structured consulting-style process: field observation, data collection, theoretical framework application, empirical analysis, and strategy formulation. Click each step to expand.
| ARCS | CIC Strength | Gap Identified | Data Signal |
|---|---|---|---|
| Attention | Curated content, feed/vlog features | Non-intuitive UI; no short-form entry; high-engagement content hidden | Top-engagement section buried at low visibility rank |
| Relevance | Some role-based content; AI trend attempts | Global content lacking; tag accuracy low; no personalized curation | AI sections: consistent high engagement, severe undersupply |
| Confidence | Campus communities; "currently watching" | AI recommendation missing; re-entry UX buried; learning path unclear | Social proof section: strong engagement confirmed at scale |
| Satisfaction | Badge system; feed sharing culture | Reward visibility near zero; no expert/creator incentives | Reading: high views / low engagement → passive consumption |
Three data-backed strategies. One phased roadmap.
All recommendations are grounded directly in the content dataset. The overarching principle: move from a view-count-driven platform to an engagement-quality-driven learning experience.
Replace view-count-only ranking with a weighted hybrid: Score = (engagement rate × 0.7) + (normalized views × 0.3). Surfaces hidden high-quality content while maintaining broad discovery. Immediately deployable without new infrastructure.
AI-related content shows ~40% higher engagement than platform average yet supply is severely limited. Expand to a dedicated AI Learning Hub, integrate real-time trend content, and position AI sections as primary entry points for upskilling-motivated learners.
Create an automated flag for content with high engagement + low views for featured placement. A "You might have missed this" section surfaces high-quality underexposed content before the view-based algorithm buries it further.
Evidence-based recommendations for a platform serving thousands of Samsung employees.
- Delivered a comprehensive diagnostic report (in Korean) revealing the views–engagement paradox — demonstrating that Samsung's curation algorithm actively surfaced less satisfying content to more users.
- Applied ARCS model across all four motivation dimensions, backed by empirical content data — producing theory-grounded, actionable recommendations rather than intuition-based suggestions.
- Proposed a hybrid recommendation algorithm as an immediately deployable fix, requiring no new infrastructure — a practical constraint given enterprise system complexity.
- Identified AI content as Samsung's highest-ROI content investment opportunity — consistent high engagement across all AI sections with severe under-supply relative to employee demand.
- Produced a phased LXP transition roadmap spanning immediate UX fixes, mid-term algorithm improvements, and long-term personalized learning path architecture.
The most counterintuitive moment came when the data showed the highest-engagement section — one I'd initially dismissed as generic — sitting at the bottom of the visibility ranking. That single finding reframed the entire project: the problem wasn't content quality, it was measurement. A platform optimizing for the wrong metric will systematically reward the wrong content. That's a learning system design problem, not a content problem — and it's exactly the gap that instructional designers are positioned to close.