How support teams can measure customer churn

Churn is usually a finance number: the share of customers who leave each month. Support teams see the reasons long before finance sees the number, in the tickets that come before a cancellation. This guide shows how a support team can measure churn in a way that points to causes: which kind of churn to count, how to split it by what happened in support, and how to tell a real difference from noise. The example numbers are invented and round, so the math is easy to follow.

Count customers and revenue separately

Our example company starts the month with 1,200 paying customers and 96,000 in monthly recurring revenue, so average revenue per account is 96,000 ÷ 1,200 = 80. During the month, 30 customers cancel and 45 join.

• Customer churn is customers lost ÷ customers at the start: 30 ÷ 1,200 = 2.5%.
• Gross revenue churn is the revenue lost from cancellations and downgrades ÷ revenue at the start: (3,600 + 1,200) ÷ 96,000 = 5%.
• Net revenue churn also subtracts the revenue existing customers added through upgrades: (3,600 + 1,200 − 2,400) ÷ 96,000 = 2.5%.

Gross revenue churn is twice customer churn here, for two reasons: the customers who left paid more than average (3,600 ÷ 30 = 120 each, against 80), and downgrades took away more revenue on top. For support, customer churn is usually the right starting point, because each lost customer is a person with a ticket history you can read. Watch revenue churn as well, so that a few large accounts leaving does not hide behind a small customer count.

Split churn by what happened in support

One churn rate for the whole company tells you how much, not why. Split the same customers into groups by their support experience in the month, and calculate churn for each group. In our example, 80 customers had a ticket reopened more than once, which often means the first answer did not solve the problem:

Other splits worth trying: customers whose question waited longer than your reply target, customers who gave a low satisfaction score, and customers who asked the same question twice. Use one split at a time, and keep the definition the same every month.

Split by how long customers have been with you

New customers often leave faster than customers who stayed through their first months, and they are the ones who ask the most setup questions. Splitting by tenure shows whether onboarding support is where the churn is:

In this example, nearly half of the lost customers (14 of 30) were in their first 90 days, although they are only a sixth of the customer base. That points to setup, and to the questions new customers ask in their first weeks.

Be careful with small groups

The reopened-ticket group has only 80 customers. One customer more or less changes its churn rate by 1 ÷ 80 = 1.25 percentage points. A difference that rests on a handful of customers can come and go from month to month by chance. Before you act on a split:

• Combine three to six months, so each group has more customers and more cancellations.
• Check that the difference appears in most months, not only in one.
• Read the tickets of the customers who left. A pattern in the tickets is stronger evidence than the rate.

Work out what is at stake

At the other group's rate of 2.14%, the 80 customers with reopened tickets would lose about 1.7 customers a month, not 6. The gap is about 4.3 customers a month.

To put a value on one customer, use customer lifetime value. With average revenue of 80 per account, a gross margin of 70% and monthly churn of 2.5%, CLV is 80 × 0.7 ÷ 0.025 = 2,240. The CLV and churn calculator works this out from your own numbers. So the gap is worth up to about 4.3 × 2,240, roughly 9,600 in lifetime gross profit for each month of customers.

Read that as the most that is at stake, not as what better support would save. Customers whose tickets get reopened may differ in other ways: they may use a harder part of the product or be unhappy for other reasons. The split shows where to look. Reading the tickets, and measuring again after you fix something, shows whether support was the cause.

Record why customers leave

Ask every cancelling customer one question with a short, fixed list of reasons and a free-text field. Keep the list stable so you can count it over time, for example:

• Too expensive
• Missing a feature
• Too hard to set up or use
• Could not get help, or help was too slow
• No longer needed, or the business changed
• Moved to another product

Then match each cancellation with the customer's tickets from the 90 days before it. "Too hard to set up" next to three tickets about the same setup step is a specific problem with a specific fix.

Signals support sees before churn

• The same customer asks the same question twice.
• A ticket is reopened, or the customer replies "that did not help".
• Questions about exporting data, downgrading or cancelling.
• A low satisfaction score on a ticket.
• A question your help pages and your support chatbot could not answer.

If you use Paperlink AI, the last signal has its own report. The Unanswered report lists the questions the bot could not answer from your documents, and Needs review collects the answers that visitors rated as not helpful. Each one is a help page to write or fix, which removes a reason for a new customer to give up.

A monthly routine

1. Calculate customer churn and revenue churn for the month.
2. Calculate churn for one support split and one tenure split, combined over the last three months.
3. Read the tickets of the customers who left in the group with the highest churn.
4. Pick one cause to fix, and note the date, so you can compare the next months with the months before.
5. Update CLV in the calculator with the new churn rate, so the value of each fix stays current.