Churn Prediction

Predict subscriber risk from behavioural signals and use the score to target follow-up or access treatment.

Business impact

• Digital subscribers — The WSJ article reports 1,389,000 digital subscribers, up from 1.08 million a year earlier, but it does not attribute that change directly to the churn model.

Data requirements

• Customer spend versus projection data (tabular) — Snowflake uses spending differences from projections to identify customers for follow-up.
• Visitor behavioural signals (tabular) — WSJ uses more than 60 signals, including frequency, recency, device, content type and location.

AI methods and techniques

• Propensity scoring — WSJ assigns a score from more than 60 signals to estimate subscription likelihood.
• Pattern analysis — The WSJ article describes using frequency, recency, device, content type and other signals to distinguish visitor groups.
• Draft generation — Snowflake's internal agent drafts follow-up questions for salespeople based on customers whose spend differs from projections.

Industries

Real-world evidence

2 documented case studies on record.

Companies using this: Snowflake, The Wall Street Journal.