Which Customers Are About to Churn and How to Stop Them (Before They Leave)

Your customer did not cancel today. They decided to cancel three weeks ago. You just found out. By the time you see the cancellation email, the relationship is already over. The real question is not "Why did they leave?" It is "Why did I not see it coming?"

Churn prediction is not a data science problem. It is a pattern recognition problem. And the patterns are visible in every digital product business — if you know where to look. This article gives you the 5 warning signals, the exact thresholds that trigger each one, the Google Sheets cohort tracker to spot trends before they become crises, and the 3-tier intervention system that turns silent exits into retained revenue.

The Hidden Cost of Silent Churn

Most digital product creators measure churn wrong. They look at monthly cancellation rate and call it a day. But cancellation rate is a lagging indicator. By the time someone cancels, they have been disengaged for weeks. The damage is done.

The real metric is silent churn — customers who have stopped engaging but have not canceled yet. They are still paying. They are still on your list. But they are already gone mentally. For subscription products, silent churn can represent 20-40% of your "active" customer base. For one-time purchase businesses, it shows up as customers who never buy again, never open emails, and never log back in.

Here is why this matters for your LTV:CAC ratio. If your true churn rate (including silent churn) is 15% but you are measuring 7%, you are overestimating LTV by 2x. You are spending money to acquire customers who will not stay. You are scaling a leaky bucket. The fix starts with seeing the leak before it becomes a flood.

The 5 Warning Signals of Imminent Churn

Each signal has a threshold, a timeline, and a specific intervention. One signal is a yellow flag. Two signals is a red flag. Three or more means the customer is 70%+ likely to churn within 30 days. Track these in your weekly dashboard alongside your core metrics.

How to Calculate Churn Rate Correctly

Before you can predict churn, you need to measure it accurately. Most creators get the formula wrong, which means they optimize for the wrong number.

The denominator matters. Use customers at the start of the period, not the average. Using the average dilutes your churn rate and hides problems. For one-time purchase businesses, measure repeat purchase rate instead: (Customers who bought again within 90 days / Total customers) x 100.

Note the one-time purchase paradox: a repeat purchase rate below 10% is actually a crisis, not a win. It means your product is not creating enough value for customers to buy again. The goal is not low churn. The goal is profitable retention that justifies acquisition spend.

Cohort Analysis: The Tool That Reveals Hidden Patterns

Cohort analysis groups customers by when they joined and tracks their behavior over time. It answers questions that aggregate churn rates cannot:

• Do customers who joined during a sale retain as well as organic customers?
• Does onboarding completion in the first 3 days predict 90-day retention?
• Which acquisition channel brings customers who stay 3x longer?
• At what month do most customers leave — the "churn cliff"?

Here is how to build a cohort retention table in Google Sheets:

The visual output is a retention matrix. Each row is a cohort. Each column is a month since acquisition. The cells show the percentage of that cohort still active. A healthy product shows a curve that flattens — most customers leave early, then the remaining cohort stabilizes. A broken product shows a straight diagonal line — customers leave at a constant rate forever.

Look at the pattern. The March cohort has 85% Month 1 retention — much higher than January's 78% and February's 72%. What changed in March? Maybe you launched a new onboarding sequence. Maybe you changed your pricing. Maybe you started requiring email verification. Cohort analysis turns "something is different" into "this specific change caused this specific result."

The 3-Tier Intervention System

Knowing who is about to churn is useless without a system to stop them. The 3-tier intervention system matches the urgency of the signals to the intensity of the response. It is automated where possible, personal where necessary, and always focused on value delivery, not retention begging.

Building the Churn Prediction Dashboard in Google Sheets

You do not need a data scientist. You need a well-structured Google Sheet. Here is the exact setup:

Sheet 1: Customer Master

• Column A: Customer ID (Stripe customer ID or email)
• Column B: Acquisition Date
• Column C: Acquisition Channel (UTM source)
• Column D: Product/Plan
• Column E: First 30-Day Login Average (baseline)
• Column F: Current 14-Day Login Average
• Column G: Core Feature Usage (last 7 days)
• Column H: Support Tickets (last 30 days)
• Column I: Last Support Ticket Date
• Column J: Payment Status (active / failed / expired)
• Column K: Email Open Rate (last 30 days)
• Column L: Signal Count (formula: counts red flags)
• Column M: Risk Level (formula: 0=green, 1=yellow, 2=red, 3+=critical)
• Column N: Last Intervention Date
• Column O: Intervention Type

Sheet 2: Weekly Ritual

• Row 1: Date of review
• Row 2: Total active customers
• Row 3: New at-risk customers (Signal Count >= 1)
• Row 4: Critical customers (Signal Count >= 3)
• Row 5: Interventions sent this week
• Row 6: Recoveries this week (customers who dropped from critical to green)
• Row 7: Churned this week
• Row 8: Net retention score (recoveries - churned)

Update this sheet weekly. The data sources are: Stripe (payment status, acquisition date), your product platform (login frequency, feature usage), your email platform (open rates), and your support tool (ticket history). Most of this can be automated with Zapier or Make.com. Even manual entry takes 20 minutes for under 500 customers.

Common Churn Prediction Mistakes

Even with the right framework, creators make predictable errors. Avoid these:

1. Tracking too many signals. Five is the maximum. More signals create noise and paralysis. If you cannot act on a signal within 24 hours of detecting it, remove it from your tracker.
2. Ignoring the first 30 days. The highest churn happens in the first month. Your onboarding experience determines whether a customer ever reaches Month 2. Cohort analysis will show this clearly — if Month 1 retention is below 60%, fix onboarding before you fix anything else.
3. Treating all churn equally. A customer who churns after 3 months is different from one who churns after 12 months. Early churn is a product-market fit problem. Late churn is a value delivery problem. The interventions are different.
4. Chasing zero churn. Zero churn is impossible and often undesirable. Some customers are wrong for your product. Let them leave. Focus on the 60% who are right for your product but are slipping away due to friction, not fit.
5. Not connecting churn to LTV. Your LTV:CAC ratio depends on churn. A 10% monthly churn rate cuts LTV in half compared to 5%. Every churn percentage point you save increases LTV by 10-15%, which means you can spend more on acquisition or take home more profit.

Connecting Churn Prediction to Revenue Forecasting

Churn prediction is not an isolated metric. It feeds directly into your 90-day revenue forecast. Here is how:

Your forecast has two components: new revenue from acquisition and retained revenue from existing customers. Most creators only model new revenue. They assume all existing customers will renew. They are wrong.

Use your cohort retention curves to model expected revenue from existing customers. If your January cohort has 52% retention at Month 6, expect 48% of their revenue to disappear by July. If your February cohort has 60% retention at Month 3, expect 40% to churn by May. Layer these expectations into your forecast and you will stop being surprised by revenue drops.

This table tells you that $3,498 in existing MRR is at risk over the next 6 months. Your acquisition target is not just "grow revenue." It is "grow revenue by enough to cover the $3,498 expected churn plus hit your net growth target." Without churn prediction, you set acquisition targets that look aggressive but actually just maintain flat revenue.

Frequently Asked Questions

What are the 5 warning signs that a customer is about to churn?

The 5 warning signs are: (1) Login frequency drops below 50% of their historical average for 14+ days, (2) Feature usage declines — they stop using core features they previously engaged with, (3) Support ticket volume drops to zero after previously being active (a silent customer is often a departing customer), (4) Payment method expires or fails without update within 7 days, and (5) NPS or satisfaction score drops below 6. Each signal has a specific threshold and timeline. One signal is a yellow flag. Two signals is a red flag requiring immediate intervention. Three or more signals means the customer is already 70%+ likely to churn within 30 days.

How do you calculate churn rate for digital products?

Churn rate = (Customers lost in period / Customers at start of period) x 100. For digital products, use a 30-day rolling window for subscriptions and a 90-day window for one-time purchases (measuring repeat purchase rate instead). The correct denominator is customers at the START of the period, not the average. For cohort-based analysis, track each month's new customers separately and measure what percentage remains active at 30, 60, and 90 days. A healthy monthly churn rate for digital products is 3-7%. Above 10% is a crisis requiring immediate product or pricing intervention. Below 2% suggests you are underpricing or not growing fast enough to test retention.

What is the best way to stop customers from churning?

The most effective churn prevention follows a 3-tier intervention system: (1) Automated re-engagement — triggered emails or in-app messages when a customer hits one warning signal, offering value (new feature, tutorial, or community access) without being salesy, (2) Personal outreach — a 2-minute personalized video or direct message when two warning signals appear, asking what changed and how you can help, and (3) Retention offer — a time-limited discount, bonus content, or extended access when three+ signals indicate imminent churn. The key insight: 60% of churn is preventable with timely intervention, but 80% of businesses never reach out until the customer has already left. The first 14 days after a warning signal appears is the intervention window. After 30 days, recovery probability drops below 15%.

How does cohort analysis help predict churn?

Cohort analysis groups customers by when they joined (acquisition month) and tracks their retention over time. This reveals patterns hidden in aggregate churn rates. For example, customers who joined during a Black Friday sale may have 40% higher churn than organic customers. Customers who completed onboarding within 3 days retain 3x longer than those who did not. Cohort analysis answers three critical questions: (1) Which acquisition channels bring sticky customers vs. discount hunters, (2) Which onboarding experiences create long-term users, and (3) At what point do most customers leave (the cliff), so you can intervene before it. The Google Sheets cohort template provided in this article automatically calculates retention curves and flags at-risk cohorts.