SVOD Subscriber Retention Survival Analysis Plan

Role: Staff Quantitative Modeler & Survival Analysis Specialist

Context

• SVOD Platform: {{svod_service_name}}
• Subscription Tier: {{subscription_tier_type}}
• Observation Window: {{observation_window_months}}
• Candidate Covariates: {{candidate_covariates}}
• Censoring Definition: {{censoring_definition}}
• Target Retention Lift: {{target_retention_lift}}

Task

Formulate a rigorous time-to-event survival analysis modeling plan for {{svod_service_name}} to quantify subscriber churn hazards, evaluate covariate effects from {{candidate_covariates}}, and support strategies to achieve {{target_retention_lift}} on {{subscription_tier_type}} accounts.

Method

1. Define the survival time random variable $T$ (tenure in days) and right-censoring indicator $\delta$ based on {{censoring_definition}} over {{observation_window_months}}.
2. Specify non-parametric Kaplan-Meier survival curves and Nelson-Aalen cumulative hazard estimators stratified by subscription cohort.
3. Formulate a semi-parametric Cox Proportional Hazards model incorporating {{candidate_covariates}}.
4. Design Schoenfeld residual test protocols to evaluate and document the proportional hazards assumption for each covariate.
5. Detail the mathematical strategy for modeling time-varying covariates (such as monthly viewing hours and catalog refresh interactions).
6. Outline parametric survival alternatives (Weibull, Log-Logistic, Gompertz) for parametric accelerated failure time (AFT) extrapolation.
7. Define discrimination and calibration evaluation protocols using Harrell's Concordance Index (C-index) and Brier scores across time horizons.

Constraints

• MUST explicitly address non-informative right-censoring assumptions and account for voluntary versus involuntary billing cancellations.
• MUST provide diagnostic remediation steps when proportional hazards assumptions are violated.
• MUST NOT treat engagement covariates as static baseline features if they fluctuate over the subscriber lifecycle.
• Restrict all explanations to statistical modeling mechanics, diagnostic checks, and tenure projections.

Output format

Provide a survival analysis technical plan structured as:

1. Event Definition & Censoring Protocol
2. Estimation Strategy (Kaplan-Meier, Cox PH, and AFT formulations)
3. Proportionality Diagnostics & Time-Varying Covariate Handling
4. Model Evaluation & Concordance Benchmarks (under 600 words)

Self-review

• Ensure censoring criteria directly integrate {{censoring_definition}}.
• Verify that each covariate listed in {{candidate_covariates}} is mapped to either a static or time-varying implementation.
• Check that C-index and Brier score evaluation windows match {{observation_window_months}}.