01 — Retention
Acquisition quality vs. durable value: a multi-method retention diagnosis
When a competitor disruption drove a 280K subscriber spike, was it real growth or borrowed users? Three independent lenses on the same question.
Survival analysis Segmentation LLM VOC
Decision impact
Reframed leadership conversation from acquisition volume to activation quality and retained LTV.
02 — Prediction
Subscriber churn prediction: survival curves and early engagement signals
Can first-month behavior separate "binge and leave" users from habit builders? Acquisition channel, billing, and engagement decomposition.
Survival curves Cohort analysis Classification
Decision impact
Surfaced channel-level churn risk and "binge and leave" patterns informing onboarding and acquisition spend.
⌘
quasi-experimental-pricing
DiD, synthetic control, RDD, and ITS on a synthetic subscription pricing dataset, with LTV translation and breakeven price calculation.
Python Causal inference DiD
⌘
transcript-analysis-pipeline
Three-stage LLM pipeline (extract, synthesize, audit) for decision-grade meeting analysis with explicit grounding and bounded fabrication controls.
Python LLM Evaluation
⌘
llm-churn-reason-mining
Pipeline that extracts and categorizes churn reasons from unstructured text using weak supervision, transformer fine-tuning, and prompt-engineered LLM summaries.
Python Transformers LLM
Series · 01
Description, prediction, and causation answer different questions
Teams confuse the three and end up making confident decisions on the wrong analytical foundation.
Series · 02
Strong teams reason about the counterfactual, not just the outcome
Leadership in data work is making counterfactual assumptions explicit, especially when you can't run an experiment.
Series · 03
Not every correlation is a lever
Teams scale the wrong strategy when they treat correlations as causal mechanisms. Signal, symptom, and lever are different.