How to Analyse Customer Churn
Churn analysis identifies who leaves, what value is lost and which causes a business can address. Begin with explicit definitions. Customer churn, gross revenue churn and net revenue churn answer different questions and should not share an ambiguous dashboard label.
Churn is most natural for an ongoing relationship, such as a subscription. For occasional purchases, define an expected return window before classifying a customer as lost.
Write a metric dictionary
For a simple monthly subscription analysis, use the customers active at the start of the month as the opening cohort. Define what makes a customer active, when cancellation takes effect and how pauses, reactivations and failed payments are treated.
| Measure | Defined calculation for the opening cohort | Question answered |
|---|---|---|
| Customer churn | Customers lost divided by opening customers | How many relationships ended? |
| Gross revenue churn | Recurring revenue lost through cancellations and downgrades divided by opening recurring revenue | How much existing revenue disappeared before expansion? |
| Net revenue churn | (Lost recurring revenue minus expansion from that cohort) divided by opening recurring revenue | Did expansion offset existing revenue losses? |
Multiply by 100 to express each result as a percentage. Exclude new-customer revenue from the revenue-churn calculations. Use the same currency basis and recurring-revenue convention throughout. These are analytical definitions; a particular billing platform may calculate its dashboard differently, so reconcile before comparing.
Hypothetical calculation
A fictional subscription business starts a month with 100 customers and $10,000 in monthly recurring revenue. Five opening customers leave. Cancellations remove $700, downgrades remove $300 and expansion from the remaining opening cohort adds $400.
Customer churn is 5 divided by 100, or 5%. Gross revenue churn is $1,000 divided by $10,000, or 10%. Net revenue churn is ($1,000 minus $400) divided by $10,000, or 6%. New customers acquired during the month do not change those opening-cohort calculations.
The cancelled accounts averaged $140 in monthly recurring revenue ($700 divided by five), compared with $100 per opening account ($10,000 divided by 100). That supports a more precise concentration finding: the customers who cancelled were larger than the opening average, while downgrades caused a further $300 of loss. None of these figures is a benchmark or a forecast. A single month also cannot establish a persistent trend.
Find where the losses concentrate
Compare cohorts at the same age. A recent cohort has had fewer opportunities to cancel than an older one. Review meaningful segments such as customer use case, plan, acquisition channel and onboarding route. Avoid splitting a tiny sample into so many groups that random variation appears to be a pattern.
Look separately at early departures and later losses. Early churn may involve poor fit, failed setup or a mismatch between sales promises and product capability. Later churn may involve reliability, changing needs or a competitor. These are hypotheses to investigate, not conclusions available from a chart alone.
Combine behavior with reasons
Use cancellation responses, support history and interviews to understand the task that failed. Let customers give an unlisted reason and record multiple contributing causes where appropriate. A forced “too expensive” category can hide a product that did not deliver enough value.
Separate voluntary departure from failed collection. Payment recovery may address an expired card, while discounts will not repair an essential missing workflow. Check whether outreach is appropriate and respect the customer’s decision to leave.
Test an intervention and follow through
Choose one cause with credible evidence. For example, if customers abandon setup while waiting for an administrator, test saved progress and clearer permission instructions. Define the eligible group, the outcome window and guardrails before release.
Track retained value and customer outcomes as well as churn. A discount that delays cancellation for one month may cost more than it preserves. Expansion can make net revenue churn negative while many customers still leave, so retain the customer and gross-loss views.
Review results after enough time for the relevant event to occur. Connect findings to onboarding and the customer feedback loop. The aim is a specific decision about a cause you can address, with a clear statement of what remains uncertain.
Reconcile the underlying customer and revenue movements
Build a simple movement table before interpreting the percentage. For each opening account, record starting recurring revenue, cancellation loss, downgrade loss, expansion and ending recurring revenue. Keep new accounts separate. Reconcile the totals to the billing records using the same period and currency convention.
Decide how to handle a customer who cancels and returns within the period. A period-end status method and an event-count method can disagree. Neither label is useful without a definition. Record the choice and keep it consistent, especially when comparing your analysis with a vendor dashboard.
Exclude one-off charges from a recurring-revenue measure unless you explicitly define another metric. Review trials, free accounts, tax, credits and refunds. An invoice adjustment can look like commercial contraction if it is mixed into the recurring base without explanation.
Translate churn into retention without changing the cohort
Under the opening-cohort conventions used here, gross revenue retention equals opening recurring revenue minus cancellation and downgrade losses, divided by opening recurring revenue. Net revenue retention also includes expansion from that same cohort. Multiply by 100 for percentages.
In the fictional example, gross revenue retention is ($10,000 minus $1,000) divided by $10,000, or 90%. Net revenue retention is ($10,000 minus $1,000 plus $400) divided by $10,000, or 94%. These complement gross and net revenue churn when all definitions match. They do not include new-account revenue.
Use the customer and revenue retention calculator to explore the declared inputs. Confirm its conventions before comparing with a billing report. A calculator can check arithmetic; it cannot decide whether the inputs represent the intended relationships correctly.
Investigate causes through a structured review
Select a sample of departures and compare the timeline: acquisition promise, setup, meaningful use, support issues, renewal and cancellation. Ask what changed and what alternative the customer chose. Preserve cases where the product worked but the need ended, since those may require a different business response from a preventable failure.
| Suspected cause | Evidence to seek | Plausible response to test |
|---|---|---|
| Setup blocker | Failed task and repeated assistance | Clearer recovery or a narrower setup path |
| Poor customer fit | Mismatch between need and supported outcome | Better qualification and offer explanation |
| Reliability issue | Incident history tied to the customer task | Correct the failure and verify recovery |
| Failed payment | Collection events and customer confirmation | Appropriate payment-update assistance |
| Changed need | Recent-event interview and current alternative | Adapt the offer only if the segment case supports it |
Treat reason codes as evidence, not a complete causal explanation. A departing customer may select the quickest option. Interviews can add context, but people may also rationalise a decision after the event. Compare accounts that stayed under similar conditions before claiming a single cause.
Prioritize one intervention and check economics
Estimate how many eligible accounts the intervention could affect and what effort it requires. Include discounts, support time, payment costs and implementation work. Retaining an unsuitable account through repeated bespoke work may consume more value than it preserves.
Avoid making cancellation harder to improve the metric. Delayed cancellation can temporarily flatter retention while increasing complaints and unresolved obligations. Give customers an understandable exit and collect feedback without making the feedback a barrier.
Review results over the relevant renewal or usage cycle. Compare cohorts at equivalent ages and record other changes that could affect the result. If a campaign coincides with a pricing change, new acquisition mix or major outage, state the attribution limit. Use the experiment plan to preserve the original hypothesis and identify what still needs investigation.