Case Studies /Before They Leave
A bank customer drifts toward the exit as warning signs go unread
Support calls and quiet accounts flag who is about to leave.Illustration · Data Stories Lab
Customer retention

Before they leave

Support calls and inactivity flag who will leave

Showing all 1,000 customers
Overview
-
customers in view
-
churn rate
-
active & high risk
-
inactive share
One in twelve customers leaves
Customers retained vs lost
Source: 1,000 card customers. Churn flag from records.
Inactive accounts carry the losses
Customers by activity, retained vs lost
Source: 1,000 card customers.
Most customers hold three or more products
Customers by number of products
Source: 1,000 card customers.
Monthly spend clusters in the low thousands
Customers by monthly spend band (MYR)
Source: 1,000 card customers.
Warning Signs
-
churn, 3+ support calls
-
churn, inactive
-
churn, first year
-
churn, single product
Churn climbs with every support call
Churn rate by number of support calls
Source: 1,000 card customers. Churn rate within each band.
Inactive customers leave about three times as often
Churn rate, active vs inactive
Source: 1,000 card customers.
The first year is the most fragile
Churn rate by tenure band
Source: 1,000 card customers.
More products, fewer leavers
Churn rate by number of products
Source: 1,000 card customers.
Customer Profiles
-
avg monthly spend
-
avg tenure (months)
-
avg products held
-
avg support calls
Risk stacks: new and single-product is the danger corner
Churn rate by tenure band, grouped by products
Source: 1,000 card customers. Bars are churn rate within each group.
Leavers and stayers spend about the same
Average monthly spend (MYR), stayed vs left
Source: 1,000 card customers. Spend is not a warning sign.
Calls and inactivity compound
Churn rate by support calls, split by activity
Source: 1,000 card customers.
Where customers sit, and who leaves
Tenure vs monthly spend, by outcome
Source: 1,000 card customers. Each dot is one customer.
Targeting & Watchlist
-
leavers in riskiest 20%
-
high-risk customers
-
leavers in view
-
lift vs working at random
Ranking by risk finds leavers fast
Share of leavers caught vs share of customers contacted
Source: risk score from the five signs. Diagonal is working at random.
Most customers score low risk
Customers by risk score band
Source: risk score from the five signs.
This month's watchlist: contact these first
Highest-risk customers in the current view (top 20)
CustomerTenure (mo)ProductsSupport callsActiveRisk score
Source: per-customer risk score. Sorted highest risk first.