# Engagement metrics that predict telecom churn risk
Sixty days before a customer calls to cancel, their app logins had already dropped by half. Their monthly data usage fell from 12GB to 4GB. And they'd contacted customer service twice in one month after nine months of silence. Nobody at the carrier noticed until the cancellation request hit the system. That two-month blind spot is the difference between a retention offer that works and a customer who's already mentally gone.
This lesson teaches you to read engagement decay as a leading indicator of churn (customers canceling or switching providers), before it shows up in the revenue numbers.
Most telecom dashboards report churn ratechurn rateChurn rate is the percentage of customers or revenue lost over a period. It measures how fast a business loses its existing customer base.Voir la définition complète →, the percentage of subscribers who leave in a given period, as a monthly or quarterly figure. In the US, postpaid wireless churn typically runs around 0.8% to 1.2% per month for major carriers like Verizon, AT&T and T-Mobile (estimate, as of 2025 investor reports). In Europe, monthly churn for mobile postpaid often sits between 1% and 1.5% depending on market competitiveness (estimate; markets like the UK and Spain run higher due to easier number portability).
The problem: churn ratechurn rateChurn rate is the percentage of customers or revenue lost over a period. It measures how fast a business loses its existing customer base.Voir la définition complète → tells you what already happened. By the time it moves, the customer has decided. Marketing and retention teams need engagement metrics, behavioral signals that predict the decision before it's final.
Telecom apps (My Verizon, MyO2, Deutsche Telekom's MeinMagenta) are used for bill checks, data monitoring and plan changes. A customer who logged in weekly and drops to zero logins for 30 days is disengaging from the brand relationship, not just the network.
Benchmark logic: compare individual login frequency against that customer's own trailing 90-day average, not against a company-wide average. A high-usage segment naturally logs in more; what matters is deviation from personal baseline.
Absolute data usage varies wildly by plan and customer type. What predicts churn is the *slope*: a sustained month-over-month decline. A customer dropping from 12GB to 4GB over two months is likely already using a second SIM, a competitor's eSIM, or Wi-Fi-only behavior signaling reduced reliance on that carrier.
This is the core of the case in the hook: usage decline started roughly 60 days before cancellation, well before any support ticket or retention flag triggered.
Two patterns both signal risk, and they're opposite:
Combine this with Net Promoter Score (NPS), a customer loyaltycustomer loyaltyYour customers' propensity to repeatedly purchase from you and resist competitive offers, driven by satisfaction, habit, trust, and switching costs.Voir la définition complète → metric based on the question "how likely are you to recommend us" scored 0 to 10, where detractors (0 to 6) after a service contact are a strong churn flag.
Individually, each signal is noisy. Combined into a weighted score, they become a usable model. A simplified version:
churn_risk_score =
0.4 * normalize(login_decline_pct) +
0.4 * normalize(data_usage_decline_pct) +
0.2 * normalize(support_contact_spike_flag)
# normalize() scales each signal 0-1 relative to
# that customer's own 90-day rolling baselineScore customers monthly. Flag the top decile for proactive retention outreach (a loyalty offer, a plan review call, a data bonus) before they self-select into the cancellation funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.Voir la définition complète →.
This is essentially a lightweight version of what carriers call a propensity-to-churn model, often built with logistic regression or gradient boosting on these same behavioral features. You don't need machine learning sophistication to start; a spreadsheet with the three signals and simple thresholds catches most of the value.
Say a mid-size regional carrier has a customer, Account X:
| Metric | 90 days ago | 60 days ago | 30 days ago | Today |
|---|---|---|---|---|
| App logins/month | 8 | 4 | 2 | 0 |
| Data usage (GB) | 12 | 9 | 6 | 4 |
| Support contacts | 0 | 0 | 1 | 2 |
Login decline: from baseline 8 to 0 is a 100% drop.
Data usage decline: from 12GB to 4GB is a 67% drop.
Support spike: from 0 to 2 contacts triggers the flag.
Even a simple rule ("flag if login decline > 50% AND usage decline > 30%") would have caught this account at the 60-day mark, two full months before cancellation, giving the retention team a real window to intervene with a targeted offer.
Once flagged, the intervention itself is a marketing decision, not just an operational one. Common tactics and their approximate effectiveness (estimates, vary heavily by market):
The CTIA (US wireless trade association) and Ofcom in the UK both publish periodic market reports with churn and switching data useful for benchmarking your own numbers against national trends.
Vérification des acquis
1. Why is churn rate alone insufficient for retention teams trying to prevent cancellations?
2. Why does the lesson recommend comparing a customer's login frequency to their own trailing 90-day average rather than a company-wide benchmark?
3. A carrier notices a customer's monthly data usage dropped sharply and they suddenly contacted customer service twice after months of no contact. How should this combination of signals be interpreted?
4. Select ALL correct answers about the difference between engagement metrics and traditional churn rate reporting.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers about why login frequency drops can signal churn risk.
Sélectionnez toutes les réponses correctes.
Not all customers behave the same way, so a single threshold across your base will misfire. Practical segmentationsegmentationDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.Voir la définition complète →:
🎬 [VIDEO: "Customer ChurnCustomer ChurnChurn rate is the percentage of customers or revenue lost over a period. It measures how fast a business loses its existing customer base.Voir la définition complète → Prediction Explained" - youtube.com - a walkthrough of how churn prediction models are built from behavioral data, useful for non-technical marketers wanting the intuition without the math]