Customer Churn Diagnostic — Contract, Lifecycle & Revenue Exposure

Contract term is the single biggest lever on churn — and roughly a third of monthly recurring revenue is currently exposed to it, concentrated almost entirely in the month-to-month base.

01Reliable

Month-to-month customers churn at 42.7% vs 2.8% for two-year contracts — a 39.9pp gap (z=29.7, p<0.001). One-year contracts sit in between at 11.3%.

Action: prioritize term-upgrade offers for the 3,875 month-to-month customers — by far the highest-risk segment.

02Reliable

The drop in churn over tenure (47.4% in months 0–12 to 6.6% in months 61–72) is largely a structural effect: the contract mix shifts in parallel, from 91.2% month-to-month at 0–12 months to 70.1% two-year at 61–72 months.

Because contract type alone separates churn by up to 39.9pp, a meaningful share of the "tenure effect" is survivorship into longer-term contracts, not loyalty deepening per customer — avoid reading the trend line alone as a behavioral maturity curve.

03Directional

Protective add-ons track with materially lower churn: no Online Security is 31.3% vs 14.6% with it (16.7pp gap); Tech Support shows a similar 16.0pp gap (both p<0.001 after BH correction across 7 comparisons).

This is a cross-sectional association, not a randomized test — the add-ons may proxy for a more engaged customer rather than causing retention. Validate with a small controlled offer test before bundling broadly.

04Reliable (about the confound)

Streaming TV/Movies and Paperless Billing show a reversed pattern — higher adoption, higher churn — purely because of confounding: streaming subscribers are 64.6% on Fiber optic vs 47.9% for non-subscribers, and Fiber optic itself churns at 41.9% vs 19.0% for DSL.

Do not target "streaming users" or "paperless-billing users" for retention outreach based on this alone — any rule built on these fields without controlling for contract/internet type will misattribute the effect.

05Directional (scenario estimate)

Churned customers represent $139.1K/month in recurring revenue (30.5% of MRR), and 86.9% of that loss comes from month-to-month accounts alone.

If month-to-month churn matched the one-year contract's 11.3% rate, the data implies roughly 1,218 fewer monthly cancellations and about $80.9K in protected monthly revenue — an illustrative upper bound, not a guarantee, since contract type is only one of several churn drivers.

Data notes & methodology (click to expand)

Definition: Churn = customer's "Churn" field is "Yes" as of the dataset snapshot. Overall churn rate = churned customers / total customers.
Shape: 7,043 rows × 21 columns. No duplicate rows and no duplicate customerID values found.
Cleaning decision: 11 records (0.16%) had a blank TotalCharges value; all correspond to brand-new accounts with tenure = 0 that had not yet been billed. These were set to TotalCharges = MonthlyCharges (their first partial bill) rather than dropped.
Limitation: this is a single cross-sectional snapshot with no calendar date field. "Tenure" lets us reconstruct a lifecycle view by grouping customers by how long they've been on the books, but it is not a true longitudinal cohort trend — each tenure bucket is a different, non-overlapping group of customers.
Limitation: the dataset has no churn-reason, competitor-switch, or price-change field, so findings here describe statistical association between account attributes and churn status, not verified causal mechanisms.
Source & license: IBM Telco Customer Churn — a publicly released sample dataset widely used for churn-analytics tutorials and education. This report is a demonstration of the analytical process, not an operational business report.

Total Customers

7,043

Single snapshot, all active + churned accounts

Overall Churn Rate

26.5%

Concentrated in month-to-month

Average Tenure

32.4 mo

Across all customers, churned + retained

Revenue at Risk

$139.1K/mo

30.5% of total MRR

1. Contract Term Is the Dominant Churn Lever

Churn rate by contract type (95% Wilson CI, baseline = Two year)

Two-prop Z-test vs. the two-year baseline for each contract type; error bars show the 95% Wilson confidence interval.

Contract	n	Churn rate	95% CI	vs baseline	Relative lift	p-value	Sig.
Month-to-month	3,875	42.7%	[41.2, 44.3]	+39.9pp	+1,408.2%	<0.001	***
One year	1,473	11.3%	[9.8, 13.0]	+8.4pp	+298.0%	<0.001	***
Two year	1,695	2.8%	[2.1, 3.7]	—	—	—	base

2. Churn Risk Across the Customer Lifecycle

Efficiency vs. structure by tenure bucket

Churn rate, %

Contract mix, % of customers in bucket

Each tenure bucket is a distinct group of customers (cross-sectional, not a tracked cohort). The top panel shows churn rate; the bottom panel shows what share of each bucket is on each contract type.

📊 Structural read: churn falls from 47.4% to 6.6% across the lifecycle, but the month-to-month share falls from 91.2% to 7.7% over the same buckets while the two-year share rises from 3.1% to 70.1%. The two panels move together, consistent with a structural (contract-mix) effect rather than a purely behavioral one — a classic setup for Simpson's-paradox-style misreading if the top panel is viewed alone.

3. Which Service Attributes Move the Needle?

Churn rate: without vs. with each service attribute

Sorted by absolute effect size. Streaming and Paperless Billing reverse direction — see the confounder note below before acting on them.

Attribute	n (without)	Churn (without)	n (with)	Churn (with)	Difference	p-value (BH-adj.)	Sig.
PaperlessBilling	2,872	16.3%	4,171	33.6%	17.2pp (▲ higher churn)	<0.001	***
OnlineSecurity	5,024	31.3%	2,019	14.6%	16.7pp (▼ lower churn)	<0.001	***
TechSupport	4,999	31.2%	2,044	15.2%	16.0pp (▼ lower churn)	<0.001	***
OnlineBackup	4,614	29.2%	2,429	21.5%	7.6pp (▼ lower churn)	<0.001	***
DeviceProtection	4,621	28.7%	2,422	22.5%	6.2pp (▼ lower churn)	<0.001	***
StreamingTV	4,336	24.3%	2,707	30.1%	5.7pp (▲ higher churn)	<0.001	***
StreamingMovies	4,311	24.4%	2,732	29.9%	5.6pp (▲ higher churn)	<0.001	***

p-values are Benjamini–Hochberg adjusted for 7 simultaneous comparisons.

⚠ Confounding warning: Streaming TV/Movies subscribers are 64.6% on Fiber optic internet vs 47.9% for non-subscribers — and Fiber optic churns at 41.9% vs 19.0% for DSL on its own. Paperless Billing customers are 62.0% month-to-month vs 44.9% for non-Paperless. These two rows reflect contract/internet mix, not a causal effect of the feature itself.

4. Payment Method Segmentation

Churn rate by payment method (95% Wilson CI)

Electronic check payers churn at roughly 3x the rate of the three automatic/paper alternatives — likely correlated with the same month-to-month, lower-commitment segment seen in Section 1.

Payment method	n	Churn rate	95% CI
Electronic check	2,365	45.3%	[43.3, 47.3]
Mailed check	1,612	19.1%	[17.3, 21.1]
Bank transfer (automatic)	1,544	16.7%	[14.9, 18.6]
Credit card (automatic)	1,522	15.2%	[13.5, 17.1]

Customer Churn Diagnostic: Contract Structure, Lifecycle & Revenue Exposure