Churn Rate Analysis: A Worked Customer-Cohort Example
Churn rate analysis measures customer or recurring-revenue losses over a defined period. Separate cancellations, downgrades, expansion, and new sales. Then investigate the underlying accounts.
A growing customer count can conceal departures. Revenue growth can also conceal customer losses when retained accounts spend more. Calculate each measure separately before deciding whether retention is improving.
The worked examples below use invented, clearly labeled data for instruction. They are not customer results, industry benchmarks, or evidence that a particular retention program works.
Define Churn Before Calculating It
For this example, a customer churns when their final paid subscription ends. The measurement period is one completed calendar month, and each account has one subscription.
Other reporting systems may recognize churn at a different point. A cancellation request today can precede the end of paid service by several weeks.
ChartMogul, for example, documents configurable churn-recognition settings. Therefore, two reports can disagree because they recognize the same event on different dates.
Write down the measurement rules before reviewing the percentage:
- What counts as one customer or account?
- Which subscription statuses qualify for the starting population?
- When does cancellation become recognized churn?
- How are failed payments, pauses, and reactivations treated?
- Which timezone and period boundaries apply?
Keep a cancellation-intent report if the team needs an earlier warning. Do not silently substitute that report for completed subscription churn.
Also distinguish customers from subscriptions. An account canceling one subscription while retaining another may represent contraction rather than a lost customer.
Customer Churn Starts with a Fixed Population
For the starting-customer method, divide lost starting customers by customers active when the period began. Multiply that result by 100 to express it as a percentage.
This follows the starting-population churn calculation. Keep both parts tied to the same population and period.
Customer churn rate = lost starting customers ÷ starting customers × 100.
Do not add newly acquired customers to that denominator. Track them separately, including any who join and leave within the month.
This example uses a fixed starting population to keep the calculation reproducible. Analytics platforms may offer other conventions, so document the method instead of comparing labels alone.
Broader SaaS revenue and adoption figures cannot supply your customer-level denominator. Your billing and account records must establish who was active at the start.
A Ten-Account Example
The following synthetic dataset represents ten accounts active on September 1. Monthly recurring revenue, or MRR, is the recurring subscription amount normalized to one month.
All amounts are US dollars. Assume no taxes, currency movements, reactivations, pauses, or multiple changes within an account during September.
Chart
A Ten-Account Example
- Starting MRR
- Ending MRR
- A
- B
- C
- D
- E
- F
- G
- H
- I
- J
- Total
| Account | Starting MRR | Ending MRR | Change during September |
|---|---|---|---|
| A | $2,000 | $3,000 | $1,000 expansion |
| B | $1,500 | $2,000 | $500 expansion |
| C | $1,500 | $2,500 | $1,000 expansion |
| D | $1,000 | $500 | $500 contraction |
| E | $1,000 | $0 | Customer churn |
| F | $1,000 | $1,000 | No change |
| G | $500 | $0 | Customer churn |
| H | $500 | $500 | No change |
| I | $500 | $500 | No change |
| J | $500 | $500 | No change |
| Total | $10,000 | $10,500 | Starting accounts only |
Accounts E and G leave, so customer churn is 2 ÷ 10 × 100 = 20%. Eight of the ten starting customers remain, giving 80% customer retention for this period.
Account D remains a customer despite spending less. Counting its downgrade as a lost customer would misstate customer churn.
Revenue tells another part of the story. The starting accounts generate $500 more MRR at month-end despite losing two customers.
That increase does not reverse those departures. It shows that expansion among retained accounts exceeds the combined revenue losses in this particular example.
Separate Gross and Net Revenue Churn
Gross MRR churn includes revenue lost through cancellations and contraction. It does not offset those losses with expansion revenue.
Using gross MRR churn, the example loses $1,500 through cancellations and $500 through contraction. Divide the combined $2,000 loss by $10,000 starting MRR to obtain 20%.
Gross MRR churn = (churned MRR + contraction MRR) ÷ starting MRR × 100.
Net MRR churn also considers expansion within the existing revenue base. ChartMogul's net MRR churn calculation also accounts for reactivation under its documented conventions.
Our example has no reactivations. Expansion totals $2,500, making net MRR churn ($1,500 + $500 − $2,500) ÷ $10,000 = −5%.
Negative net revenue churn means expansion exceeds the included revenue losses. It does not mean the company lost a negative number of customers.
The corresponding retention figures make the distinction easier to read:
| Measure | Calculation for this starting population | Result |
|---|---|---|
| Customer retention | 8 remaining customers ÷ 10 starting customers | 80% |
| Gross revenue retention | ($10,000 − $1,500 − $500) ÷ $10,000 | 80% |
| Net revenue retention | ($10,000 − $1,500 − $500 + $2,500) ÷ $10,000 | 105% |
Keep these percentages attached to their definitions. A presentation showing only 105% net revenue retention would omit the two lost accounts.
MRR is also not the same as cash received or accounting revenue recognized. Do not replace the recurring-revenue inputs with invoice collections without changing and explaining the measure.
New Sales Can Hide Existing-Customer Losses
Now add two hypothetical customers acquired during September. Assume they finish the month with $1,200 and $800 MRR respectively, without churn or other changes.
The business starts with ten customers, loses two, and adds two. Its ending customer count is ten, unchanged from the beginning.
Yet customer churn remains 20% under the starting-customer method. Comparing only opening and closing totals would miss both departures.
Total ending MRR becomes $12,500 after adding $2,000 from the new customers. Only $10,500 belongs to the original ten-account population.
Reconcile the movement before presenting growth:
- Begin with $10,000 starting MRR.
- Subtract $1,500 churned MRR and $500 contraction.
- Add $2,500 expansion from starting customers.
- Add $2,000 new-customer MRR separately.
- Confirm $12,500 total ending MRR.
Do not insert the new $2,000 into the net retention numerator. That would mix customer acquisition with retention of the existing revenue base.
Use Cohorts to Find When Losses Occur
A cohort groups customers sharing a starting characteristic, such as their first paid month. Following each group at the same age can reveal patterns that calendar totals obscure.
The next table is a separate synthetic dataset. It assumes each cohort begins on its month's first day, with no reactivations.
The observation cutoff is September 30. Every percentage uses that row's original customer count as its denominator.
| First paid month | Starting customers | End of first month | End of second month | End of third month |
|---|---|---|---|---|
| July | 20 | 18, or 90% | 16, or 80% | 15, or 75% |
| August | 30 | 27, or 90% | 24, or 80% | Not yet observed |
| September | 25 | 20, or 80% | Not yet observed | Not yet observed |
Compare September's first-month retention with the earlier cohorts at that same age. Do not compare it with July's third-month result as though the observation windows matched.
The table suggests a question about September's early experience. It does not establish whether onboarding, acquisition mix, pricing, or another factor caused the difference.
Missing future observations are not zero retention. Mark them explicitly so a chart does not imply that customers disappeared before observation was possible.
Small populations also deserve careful interpretation. One additional departure among twenty starting accounts changes the rate by five percentage points.
Segment Without Changing the Question
Choose segments that can support a practical response. Starting plan, acquisition source, contract type, and customer age may help identify where investigation should begin.
For retention by starting plan, assign customers to their plan at the period's beginning. Preserve that assignment when an account later upgrades or downgrades.
Otherwise, a customer moving between plans can appear to leave one population without leaving the business. Your segment report needs to distinguish that movement from company-wide customer churn.
Record the denominator beside every percentage. Losing one of two customers requires different interpretation than losing 500 of 1,000.
Avoid averaging segment percentages without weighting their starting populations. Recalculate the combined rate from the combined counts or revenue amounts.
For example, one loss among ten customers and nine among ninety both produce 10% churn. Other combinations can make an unweighted percentage misleading when segment sizes differ.
The ownership and relationship choices within a CRM operating model can affect these joins. Confirm which account identifier connects billing records, customer success notes, and sales history.
Separate Observed Reasons from Assumptions
A cancellation survey records what a respondent reports. It does not necessarily explain every departure or establish the effect of a proposed fix.
Combine that feedback with relevant account history. Check onboarding completion, support cases, usage changes, contract events, and billing outcomes where available.
Separate voluntary cancellation from payment-related loss. Stripe documents payment recovery tools, including retries and customer communications, for failed recurring payments.
A failed payment is not automatically final churn under every recognition policy. Confirm the subscription's subsequent status and recovery outcome before classifying the account.
Structure an investigation around evidence:
- Observed pattern: which segment or cohort changed?
- Supporting records: which accounts and events explain the calculation?
- Hypothesis: what might explain the pattern?
- Proposed response: what change will be tested?
- Evaluation: which future measure and comparison will assess the result?
Keep “reason unknown” available instead of forcing every customer into a convenient explanation. Missing evidence is better recorded than replaced with certainty.
Turn the Analysis into a Repeatable Review
Begin each review by reconciling the starting population, movements, and ending balances. Investigate differences before discussing whether the percentage is acceptable.
Use a consistent definition across periods. If a recognition setting changes, explain the change and assess whether earlier periods need recalculation.
Do not adopt a universal “good churn” threshold without a comparable population. Contract duration, account value, customer age, and measurement conventions can change what a comparison means.
The same care used in interpreting software buyer research applies to retention benchmarks. Check who was measured and how before applying the result to your business.
A spreadsheet may support an initial reconciliation. For recurring reporting, evaluate ChartMogul or Baremetrics against the same account-level calculation.
Check integrations, recognition settings, exports, and handling of edge cases before adopting either tool. Automation should reproduce an agreed definition rather than conceal an unexplained difference.
End the review with a named owner and one testable follow-up. In the sample, investigate the two departed accounts while separately reviewing the durability of expansion revenue.
The objective is a decision supported by traceable movements. A favorable headline percentage cannot replace that explanation. Repeat the churn rate analysis each period with the same definitions so the trend stays comparable.
Frequently Asked Questions
How Do You Calculate Customer Churn Rate?
Divide lost starting customers by the starting customer population, then multiply by 100. State the period and recognition policy alongside the result.
Can Revenue Grow While Customers Churn?
Yes, expansion or new sales can outweigh revenue lost from departing customers. Keep new sales outside existing-customer retention calculations.
What Does Negative Net Revenue Churn Mean?
Expansion and any included reactivation exceed the revenue losses counted in that measure. It does not mean that no customers left.
Is Monthly Churn Multiplied by Twelve the Annual Rate?
Not generally; that ignores compounding and changing populations. Calculate annual churn from the appropriate annual starting population and observed losses.
Resources
- Stripe cohort analysis: grouping customers and comparing retention over time.
- ChartMogul churn recognition: timing and configuration differences.
- ChartMogul MRR movements: contraction, churn, and subscription changes.
- Stripe subscription analytics: subscription metrics and configurable definitions.
- Stripe revenue recovery: handling failed recurring payments.
- Baremetrics subscription analytics: product information for an optional reporting evaluation.