Sarah runs a SaaS company with 200 customers. Every month, 4-6 of them cancel. Her churn rate is 2.5% monthly — 26% annually. At her average contract value of $800/month, that's $192K in annual revenue walking out the door.

The worst part isn't the money. It's the surprise. Every cancellation feels like it came out of nowhere. "We loved the product," they say on the exit survey. "We just don't have the budget right now."

But when Sarah looks back at the data, the signals were there weeks before the cancellation. The customer's login frequency dropped. Their support tickets increased. Their usage of key features flatlined. The data told the story — nobody was listening.

This is the churn prediction gap. Your tools have the data. But unless someone (or something) is monitoring every customer across every data source continuously, the signals get missed.

The 7 Churn Signals Hiding in Your Data

What to track: Weekly active sessions per customer compared to their baseline (average over the first 3 months).

The threshold: When a customer's login frequency drops below 50% of their baseline for 2+ consecutive weeks, they're at risk.

Why it matters: A customer who logged in daily but now logs in twice a week isn't "busy." They're disengaging. This is the earliest and most reliable churn signal — it typically appears 30-45 days before cancellation.

Signal 2: Dropping Feature Usage

What to track: Usage of your "sticky" features — the ones that correlate with long-term retention.

The threshold: When a customer stops using 2+ key features they previously used regularly.

Why it matters: Every SaaS has 2-3 features that define whether a customer is getting value. If they stop using those, they've mentally moved on even if they're still paying.

Signal 3: Rising Support Ticket Volume

What to track: Support tickets per customer per week, indexed to their historical average.

The threshold: 3x increase in ticket volume within a 2-week window.

Why it matters: Happy customers don't contact support. A sudden spike means something is broken in their workflow, and if it doesn't get fixed fast, they'll find a tool that works better.

Signal 4: Payment Failures

What to track: Failed payment attempts, declined cards, expired payment methods.

The threshold: Any payment failure on an account that hasn't been resolved within 48 hours.

Why it matters: Payment failures drive 20-40% of all subscriber churn. Some customers don't even know their payment failed — they just drift away. Others use it as a convenient exit point.

Signal 5: Contract Approaching Renewal Without Expansion

What to track: Accounts within 60 days of renewal that haven't expanded (added seats, upgraded plans, increased usage).

The threshold: Any account approaching renewal with flat or declining metrics.

Why it matters: Healthy accounts grow. If a customer has been on the same plan for 12 months with no expansion, they're getting minimum value — and are vulnerable to a cheaper competitor.

Signal 6: Champion Departure

What to track: When your primary contact (the person who bought your product) leaves the company or changes roles.

The threshold: LinkedIn job change, email bounces, new contact requesting admin access.

Why it matters: Most B2B purchases are champion-driven. When the champion leaves, the new person inherits a tool they didn't choose and may not understand. This is the highest-risk churn event — it typically leads to cancellation within 90 days unless you re-onboard the new contact.

Signal 7: Competitive Activity

What to track: Customers visiting competitor comparison pages, searching for alternative tools, or mentioning competitors in support conversations.

The threshold: Any combination of competitor research + another churn signal.

Why it matters: A customer researching alternatives while their usage is declining is not casually browsing. They're actively evaluating a switch. This is the most urgent signal — you may have days, not weeks.

The Churn Prevention Playbook

Detecting signals is half the problem. The other half is acting fast enough to matter.

Tier 1: Automated (No Human Needed)

Failed payment → automated retry sequence with smart timing based on historical success rates
Approaching renewal → automated check-in email from the account owner, not marketing
Usage drop → in-app message highlighting underused features with a tutorial link
Expiring payment method → proactive alert to the customer 30 days before expiration

Tier 2: Human Touch (High-Value Accounts)

Support spike → account manager call within 24 hours for accounts above $5K ARR
Champion departure → executive outreach to the new contact with a personalized re-onboarding offer
Multiple signals firing → save offer (extended trial, temporary discount, dedicated support session)

Tier 3: Strategic (Systemic Fixes)

Recurring support issues → product fix (if 10+ customers hit the same issue, it's not a support problem — it's a product problem)
Feature adoption gaps → onboarding improvement (if customers aren't using sticky features, your onboarding isn't teaching them)
Consistent churn at month 3 → activation audit (something in the first 90 days isn't delivering value fast enough)

Building the Early Warning System

Option 1: Manual (Works Until ~50 Customers)

Create a spreadsheet with:

Customer name, ARR, renewal date
Weekly login count
Monthly support ticket count
Last feature usage date
Payment status

Review weekly. Flag any customer with 2+ declining metrics. Time investment: 2-3 hours/week.

Option 2: Connected (50-500 Customers)

Connect your product analytics (Mixpanel), support tool (Intercom/Zendesk), and payment processor (Stripe) to a single platform that:

Monitors all signals continuously
Scores each customer's health (0-100) based on weighted signals
Alerts you when a customer crosses from "healthy" to "at risk"
Recommends specific actions based on which signals triggered

This is what NuMoon's churn prediction module does — it connects your tools, monitors the signals, and tells you exactly which customers need attention and why.

Option 3: Fully Automated (500+ Customers)

At scale, you can't manually review every at-risk account. You need:

Automated tier-1 interventions (retry payments, send emails, trigger in-app messages)
Automatic escalation to account managers for high-value accounts
AI-generated health scores that update daily
Predictive models that learn from your historical churn patterns

The Math: Why Churn Prevention Is Your Highest-ROI Investment

Reducing churn by 5 percentage points (e.g., from 8% monthly to 3% monthly) has a dramatic compounding effect:

| Annual Churn Rate | 3-Year Customer Lifetime Value (at $800/mo) | |---|---| | 30% (2.5%/mo) | $19,200 | | 15% (1.25%/mo) | $38,400 | | 5% (0.4%/mo) | $76,800 |

Cutting your churn rate in half quadruples your customer lifetime value over 3 years. No marketing campaign, no new feature, no pricing change has this kind of impact on unit economics.

Frequently Asked Questions

What's a "good" churn rate?

For B2B SaaS: 3-5% annual gross churn is excellent, 5-10% is acceptable, above 10% is a problem. For B2C/SMB: 5-8% monthly is common, below 3% monthly is strong. Compare yourself to companies at your stage and price point, not mature enterprises.

How early can you predict churn?

With connected data across product usage, support, and payments, you can typically identify at-risk customers 30-45 days before they cancel. The earlier you catch it, the higher your save rate — accounts flagged 30+ days early have a 40-60% save rate. Accounts flagged the week before renewal have less than 10%.

Should I offer discounts to prevent churn?

Only as a last resort, and only if the reason for leaving is genuinely price-related. If the customer is leaving because the product doesn't deliver value, a discount just delays the inevitable. Fix the value problem first. If price is the real issue and the customer is otherwise healthy, a 3-month discount can buy time while you demonstrate ROI.

What if I don't have product analytics?

Start with what you have. Payment data (from Stripe) gives you signal 4 and 5. Support data (from Intercom/Zendesk) gives you signal 3. Even without product analytics, these two sources catch 40-50% of at-risk accounts. Add product analytics when you can for the full picture.

Every Churned Customer Is a Failure to Listen

The data is there. Your Stripe account knows who failed a payment. Your Intercom knows who submitted 5 tickets this week. Your Mixpanel knows who stopped using the core feature.

The problem isn't data. It's connection. These signals sit in separate tools, monitored by different teams (or nobody), with no system linking them to a single customer health score.

Connect your tools. Monitor the signals. Act before the customer decides to leave.

Take the free health scan to see how connected your customer data really is.