From Generic Carousels to Adaptive Journeys
PS Plus has rich catalogs, but carousels are often static. AI-driven ranking can tailor picks per cohort and session intent.
Signals That Matter
- Session intent: Quick play vs. weekend deep-dive.
- Device context: Console vs. remote play; controller vs. handheld.
- Social graph: Party history, co-op preferences, voice channel proximity.
- Catalog state: New drops, expiring titles, genre gaps.
Model Architecture
- Retrieval: Two-tower model with embeddings for users, games, and sessions.
- Rerank: XGBoost/LightGBM with price sensitivity, backlog pressure, and time-to-fun.
- Explore/Exploit: Thompson sampling to safely test new titles.
Churn-Aware Offers
LTV and churn models gate discounts and trial lengths. High-risk users see tailored win-back offers; low-risk users see discovery-focused surfacing.
Safety, Trust, and Controls
- Age gating and content ratings respected in ranking.
- Explainability snippets: “We picked this for your co-op streak.”
- Opt-outs and data portability for regulatory compliance.
Impact
Expect higher session starts, lower churn, and better attachment to PS Plus Extra/Premium without overwhelming players.