Churn analysis for B2B: formulas, cohort cuts and churned-customer interviews

Churn analysis answers two questions: how much are you losing, and why? The first half is arithmetic on your billing data: customer churn rate, gross revenue retention and net revenue retention, then the same numbers cut by cohort, segment and tenure until the losses concentrate somewhere. The second half is talking to the customers in that concentration, because billing data shows who left and when, but almost never the reason.

This guide covers both halves for a B2B subscription business, with formulas and a worked example you can copy into a spreadsheet.

A worked example

Suppose you sell scheduling software to field service companies (HVAC, plumbing, electrical). All figures here are hypothetical. At the start of the quarter you have 200 customers paying $100,000 in monthly recurring revenue (MRR). During the quarter:

• 12 customers cancel. Together they paid $4,800 a month.
• Some customers downgrade, removing $1,200 of MRR.
• Other customers add seats, adding $5,000 of MRR.
• You also sign 25 new customers. Leave them out of every churn calculation below; churn measures what happened to the customers you started the period with.

The three formulas to calculate first

These definitions follow ChurnZero's churn rate glossary and its pages on gross and net revenue retention.

Customer churn rate = customers lost in the period ÷ customers at the start of the period

12 ÷ 200 = 6% for the quarter.

Gross revenue retention (GRR) = (starting MRR − churned MRR − contraction MRR) ÷ starting MRR

($100,000 − $4,800 − $1,200) ÷ $100,000 = 94%. Gross revenue churn is the other side of the same number: 6%.

Net revenue retention (NRR) = (starting MRR + expansion MRR − contraction MRR − churned MRR) ÷ starting MRR

($100,000 + $5,000 − $1,200 − $4,800) ÷ $100,000 = 99%.

Look at the three together. NRR of 99% looks almost flat, yet 6% of customers left. MetricHQ's comparison of NRR and GRR makes the same point: GRR excludes expansion and cannot exceed 100%, so it shows retention problems that growth from other accounts can mask in NRR. Customer count and revenue can also diverge. ChurnZero's example shows that losing higher-paying customers makes revenue churn higher than customer churn, and losing lower-paying ones does the opposite.

Two rules keep the numbers comparable over time: use the same period length every time (monthly or quarterly), and decide once how to treat customers who cancel and come back.

Cut the numbers until the losses concentrate

A company-wide churn rate rarely tells you what to do. Split it along a few lines and look for the group where losses cluster.

By cohort. Group customers by the quarter they started and track how many are still paying after the same number of months.

If newer cohorts retain worse, something changed around the time they joined: a new acquisition channel, a pricing change, a different onboarding process or a different kind of customer.

By segment. In the example, suppose 90 of the 200 customers have fewer than 10 technicians and 110 have 10 or more. If 10 of the 12 cancellations came from the smaller group, that is 10 ÷ 90 = 11.1% churn for small accounts versus 2 ÷ 110 = 1.8% for larger ones. Other useful segment cuts: plan, industry, region and the sales channel that brought the customer in.

By tenure at cancellation. Customers who leave in the first few months often never got set up properly. Customers who leave after two years more often had something change: a new owner, a new competitor offer, a price increase.

By early behavior. If you track usage, compare accounts that churned with those that stayed on a few early actions, such as whether they finished setup or how many technicians logged in during the first month.

Keep the sample size in view. Ten cancellations out of 90 is a pattern worth investigating, but one or two more or fewer would change the rate noticeably. Treat small-group differences as leads for interviews, not conclusions.

Use cancellation reasons as a starting point

Most billing and CRM tools let customers or account managers pick a cancellation reason. These are useful for sorting, but they are short and often chosen to end the conversation quickly. "Too expensive" can mean the customer never used the product enough to justify the price, a competitor offered a discount, or the budget owner changed. You will not know which without asking.

Interview churned customers where the numbers point

The analysis should tell you whom to call. In the example, you would start with small field service companies that cancelled within their first six months, not a random sample of everyone who left. Aim for a handful of conversations in that group, then decide whether you need more.

Ask each person to walk you through the account's history: why they signed up, what setup looked like, the first time something did not work, what they tried, and what finally triggered the decision. Then ask what they use now. MaRS's win-loss guide points out that a churn interview for you is often a win interview for a competitor, so what they moved to matters as much as what they left.

Our guide on how to interview churned customers covers the call itself, including how to separate the trigger from the underlying problem. How to check customer interviews against product data covers bringing the conversations back to your numbers.

Turn the findings into one change and one metric

End each round with a short write-up: the group where churn concentrates, what the interviews showed, the one change you will make, and the number you expect it to move. In the example, if interviews show small companies never connected their dispatch calendar, the change might be a guided setup step, and the metric would be six-month retention for the next small-account cohort. Recalculate after one full cohort has passed through.

Your next step

Pull last quarter's customer list with start dates, MRR and cancellations, and calculate the three formulas above. Then make the cohort and segment tables and pick the group with the worst retention. If the people you need to talk to are hard to reach, or you want to hear from operators who chose a competitor from the start, search for people by describing their role and company type, or start from the operations professionals in facilities management directory page. Instant Expert finds people whose work matches your question. You review them, it sends the invitations, and you pay for each call that gets booked.