Customer Attrition Predictor

Flags retail customers likely to move their main account.

Strictly ConfidentialBanking secrecyCustomer PIILimited riskScoring · 2.1.0 · customer-attrition-predictor

Test in Sandbox

88 41K/30d 46

AUC (out-of-time)

0.812

Lift in top decile

3.9x

Precision@top 5%

0.31

PSI (last month)

0.06

About this model

Scores every active retail customer monthly on the likelihood of closing their main account or moving salary credit elsewhere in the next 90 days. Surfaces the top drivers per customer (falling balances, fewer salary credits, recent unresolved complaint, card dormancy) so retention teams know why, not just who.

Intended use

Prioritising retention outreach by relationship managers and the contact centre, and sizing retention campaigns. Not for pricing, credit or any adverse decision about a customer.

Training data lineage

Rebuilt from the SAS Enterprise Miner churn model used by Consumer Banking since 2020. Features built in BigQuery from customer-360-mart, casa-transactions (salary-credit and balance trends) and complaints-corpus (open complaint flags). Trained on 30 monthly snapshots in Vertex AI Pipelines; registered in the Vertex AI model registry after Model Risk Management confirmed parity with the SAS model.

Limitations & bias notes

Less reliable for customers with under six months of history. Salary switching to e-wallets is only partly visible. Scores drift after large repricing events and should be read alongside campaign calendars.

#attrition#churn#retention#consumer-banking#xgboost#rebuilt-from-sas

Ownership and sensitivity

Owned by

Consumer Banking Customer Analytics & CRM

Accountable owner: Mei Ling Tan

Who approves access

Owner, Customer Analytics & CRM, Mei Ling Tan
Data Governance Office, Ahmad Faizal

In Grow the customer book, Rebuilt from SAS

Strictly Confidential

Customer-level data under banking secrecy and PDPA, and models that decide on customers.

Named-user entitlement with the data owner's consent and a stated purpose; reviewed on expiry.

Banking secrecy: Customer account information protected under the Financial Services Act 2013.
Customer PII: Contains or processes personal data about customers (PDPA 2010).

Try it

Live sandbox

customer=C-00311852&tenure_years=11

fixture

Evaluation only: restricted models run against synthetic or masked sample data until your business unit's access request is approved.

Feedback

Details

Updated: 2026-07-02
Latest version: 2.1.0
Licence: BU Restricted
Access: Restricted
Framework: XGBoost
Language: N/A

Trained on

customer-360-mart →casa-transactions →complaints-corpus →