Churn Analysis: Find the Cause Before Sending More Emails

Churn analysis identifies which customers leave, when they leave and what evidence explains the departure. A single cancellation rate tells you the size of a problem. It cannot tell you whether the right response is better qualification, a simpler first workflow, a product fix or a payment follow-up.

For a B2B team, this distinction changes outreach. Sending every departing customer another offer treats different problems as one campaign. Start with a defined population, compare meaningful groups and inspect the reasons before deciding whom to contact. The workflow below combines two operators’ podcast lessons with explicit measurement rules and a fictional example.

What churn analysis should answer

A useful analysis connects three questions: who left, what happened before the departure, and what the team can change. Customer churn counts lost customer relationships. Revenue churn measures recurring revenue lost through cancellations or reductions. Neither is interchangeable with a quiet user, a missed meeting or a lead who never purchased.

Choose a definition that matches the business. For a subscription product, an account might count as lost when its paid service ends. A cancellation request and the final service date may differ. For a project business, an inactive customer does not automatically represent a cancelled recurring relationship. Write the rule before extracting records.

This guide focuses on existing B2B customers. Lead qualification happens before purchase; customer onboarding helps a new customer reach value. Churn analysis examines departures across the customer base and can reveal weaknesses in either process.

Two podcast lessons: diagnose the group, then inspect the work

In Startups for the Rest of Us episode 838, Rob Walling describes asking founders who is churning, whether the customers fit their intended profile, and why they leave. He also recounts examining where a demanding onboarding process loses people. The operational lesson is to locate the failure before choosing a remedy.

Walling separately discusses a low-priced plan with different retention economics from higher-priced plans. That example is a reason to segment the investigation. It does not establish that raising prices improves retention in every company, or that every low-priced customer is a poor fit.

In The SaaS Podcast’s interview with Ron Hash, the Skimmer founder describes several changes alongside improving retention: supporting phones customers already owned, welcome calls, service and a focused onboarding flow. That flow helped pool-service businesses enter real customers, prepare the next route and use the product in the field.

Hash’s account illustrates a possible activation problem hidden inside a churn number. It is a retrospective account with several changes, not a controlled test of one email or checklist. The method below is our editorial synthesis; neither guest presented this exact workflow.

Step 1: define the population and the loss event

Record the analysis period, account unit, active-customer rule, opening population and effective departure date. Keep a stable customer identifier so a changed email or company name does not become a false cancellation. Reconcile billing records with contract and delivery information where necessary.

For a clearly defined opening population, customer churn is the number of those customers lost during the period divided by the number present at its start, multiplied by 100. Stripe’s measurement guide describes this common starting-population convention. Document how new customers, reactivations and account mergers are handled instead of silently changing the denominator.

Compare like periods. An annual contract with no renewal due is not exposed to the same decision as an account whose contract expires this month. Show renewal eligibility when that is the relevant risk set. Keep this analysis separate from a monthly whole-base rate, and label both denominators.

Step 2: separate customer loss from revenue loss

A lost small account and a lost large account each count as one customer departure. Their revenue consequences differ. Keep account counts and recurring revenue beside each other, with a consistent opening population.

Gross revenue churn includes recurring revenue lost to cancellation and contraction; net revenue churn also subtracts expansion from that same existing population. Stripe’s revenue churn explanation distinguishes these measures. New-customer revenue does not become expansion of the opening customer base.

For example, a fictional business starts with $20,000 in monthly recurring revenue. It loses $1,000 through cancellations and reductions, and gains $1,500 through expansion. Gross revenue churn is 5%; net revenue churn is minus 2.5%. Growth from retained accounts can coexist with real departures. A reassuring net number should not erase the accounts that left.

Step 3: segment by a question you can act on

Begin with a few dimensions supported by records: customer fit at purchase, tenure, plan, acquisition source or progress toward the agreed first result. Choose the grouping because it can change a decision. Do not create dozens of tiny segments until a dramatic percentage appears.

Compare groups with similar exposure and follow-up. A recently launched cohort has had less time to cancel than an older cohort. An enterprise renewal should not be compared directly with a monthly self-service subscription. State the counts as well as percentages, and keep incomplete observations visible.

Fit labels also need discipline. Do not redefine every departed customer as outside the ideal customer profile after the fact. Use the criteria recorded at acquisition, or mark a retrospective assessment explicitly. Our ICP guide provides the upstream context; the analysis should test that context rather than protect it from disconfirming evidence.

Step 4: keep the reason and the evidence separate

Create a compact departure record: account, effective date, segment, observed first-value milestone, stated reason, supporting source, unresolved question and responsible owner. Store the customer’s explanation separately from the team’s interpretation.

A customer might say the product costs too much. That is a stated reason. Whether the deeper issue was missing value, a budget change or a cheaper alternative remains unconfirmed until additional evidence supports it. Likewise, inactivity can justify a check-in; it does not prove dissatisfaction.

Useful evidence can include a cancellation response, an agreed implementation milestone that remained blocked, a documented missing requirement or a failed payment recorded by billing. An empty reason field should remain unknown. AI can organize approved notes and flag contradictions, but it should not complete a missing explanation with a plausible story.

A fictional example: the same rate, two different actions

Consider 100 paying accounts present at the start of a month. Twenty are still in their first implementation stage; eighty have been customers longer. By the end, six from the first group and two from the second have left. Overall customer churn is 8%. Within the groups, the observed rates are 30% and 2.5%.

Those counts identify a place to investigate, not a proven cause. Suppose records show that four of the six early departures never completed the first real workflow. Two others provide no reason. The right next step is to inspect the blocked workflow and speak with willing customers, while retaining the unknowns.

Now suppose the two established accounts ended service after their projects closed. Treating all eight departures with the same discount would ignore these distinctions. The early group might need a narrower implementation step; the established group might offer feedback about contract fit. These are hypotheses to verify, not reported Skimmer results.

Step 5: turn the finding into one bounded intervention

Write an intervention card with the affected group, evidence, proposed change, owner, start date and review condition. If first value is blocked, improve that specific dependency. If the original sale promised an unsupported use case, correct qualification and the handoff. If billing shows a failed payment, route the case to the payment process.

Contact customers about the problem you can help solve. A short message might ask whether the missing import is still blocking the agreed workflow and offer a concrete next step. Do not claim to know why a customer left when the record contains only an inference.

Use your CRM to preserve verified fit, source links and ownership; keep contract and billing measures in the appropriate systems. A sales account plan can coordinate work for a strategic customer. It should not replace the cohort ledger or turn every customer into a forecasted rescue opportunity.

Review outcomes without claiming causation too early

After the change, compare cohorts with similar tenure, fit, contract terms and observation windows. Track the targeted operational milestone alongside customer and revenue losses. A completed workflow is an intermediate result; it does not guarantee renewal.

A smaller churn rate after an intervention may also reflect seasonality, a different customer mix or fewer renewals due. Preserve those alternatives. For a small B2B base, individual departure reviews may teach more than unstable percentages. Report counts, unknown reasons and incomplete periods before declaring success.

The practical output is a decision: a specific group, a supported explanation or an honest unknown, and the next evidence-producing action. More outreach becomes useful when that decision gives it a purpose.