Churn propensity scores

Detects churn propensity score outputs: a churn propensity / churn score / attrition risk phrase combined with populated scoring data - rows keyed by a customer identifier with a score, labelled churn scores or risk bands, or labelled customer/subscriber identifiers. Phrase-only mentions stay discovery-only.

Type: regex
Engine: boost_regex
Confidence: medium
Confidence justification: Enforce tiers require populated dataset structure (a data row keyed by a customer/record identifier, a labelled value or a labelled identifier) beside the topic phrase; a column header on its own only counts toward 85, so blank extracts stay at 65. Help pages, policies, product copy and news that only name the topic stay at 65. Medium because internal analytics documentation with worked examples can still reach 75.
Detection quality: Topic false positive
Jurisdictions: global
Regulations: Privacy Act 1988 (Cth), Information Privacy Act 2009 (Qld), NDB Scheme (Cth), GDPR
Frameworks: ISO 27001, ISO 27701
Data categories: pii
Scope: wide
Risk rating: 8
Platform compatibility: Purview: Compatible, GCP DLP: Compatible, Macie: Compatible, Zscaler: Compatible, Palo Alto: Degraded, Netskope: Unsupported

Pattern

(?i)\b(?:churn\s+(?:propensity|prediction|risk|probability|likelihood)(?:\s+scores?)?|churn\s+scores?|propensity\s+to\s+churn|attrition\s+(?:risk|propensity|scores?)|likelihood\s+to\s+(?:churn|leave|cancel)|retention\s+model\s+(?:scores?|outputs?))\b

Corroborative evidence keywords

retention offer, save desk, win-back, contract end, model version, top decile, feature importance, next best action, discount offer

Proximity: 300 characters

Should match

Churn propensity scores - model v4.2 scored 2026-09-01 customer_id,plan,tenure_months,churn_score,risk_band 50018821,Mobile 40GB,14,0.87,High 50018840,NBN 100,48,0.12,Low Top decile routed to the save desk for a retention offer. — Scoring extract - header, customer rows and retention evidence (85)
Churn risk review Customer ID: 50018821 Account flagged by the retention team. — Single labelled customer identifier (75)
Save desk call card - attrition risk alert Subscriber number: S88104427 Churn probability: 0.91 Risk decile: 1 Contract end: 30/11/2026 (Home Broadband 250) Next best action: offer 6-month $15 discount offer; escalate to win-back team if declined. — Call card with subscriber number, labelled churn probability and decile plus retention-action evidence (85)

Should not match

Fact sheet: a churn propensity score estimates how likely a customer is to leave. Scores range from 0 to 1 and are refreshed monthly; see the model card for features. — Model explainer
Churn scores customer_id,plan,tenure_months,churn_score,risk_band — Blank extract - column header only (65)
Telcos are using churn prediction models to cut customer losses, with churn rates falling to 1.1% a month, a news report said. — News item
unrelated generic text without domain phrases — Unrelated
customer invoice — Generic word pair

Known false positives

Privacy notices, help-centre articles, product pages and policies describing churn models and scores. Mitigation: They name the topic without populated rows, labelled values or identifiers, and guidance/help framing is excluded, so they stay at 65.
Analytics training material and documentation with worked examples of churn models and scores. Mitigation: Training and template framing is excluded; unframed worked examples with realistic rows can still reach 75.
News and industry commentary with aggregate statistics. Mitigation: Aggregate percentages are not keyed to a record identifier, so they do not satisfy the row or labelled-value structures.