How to measure churn in a fast-growing company

Short answer

How should churn be measured when a company is growing fast?

Measure churn among the customers and revenue you had at the start of the period, not against a customer base inflated by new sales. Weight by revenue to see the financial effect, and by exposure time if you have to include new customers. Follow cohorts, and separate new customers from established ones – otherwise changes in the customer mix will look like changes in churn.

The churn percentage looks like a simple figure: the customers who left, divided by the customers you have. In a stable business that works reasonably well. In a company where the customer base and revenue grow every month or quarter, the same formula can produce figures that are systematically too low, systematically too high, or that change for reasons that have nothing to do with customer loyalty.

This article explains why, and how to measure churn so that the figure can actually be used to steer by. Basic concepts such as customer churn, revenue churn, GRR and NRR are explained in How do you measure the value of Customer Success?

Why growth distorts the churn figure#

Three mechanisms make growth tricky:

1. The denominator grows faster than the risk. Customers who arrived during the month have barely had time to become customers. If you include them in the denominator, the churn percentage is lower than the real risk for the customers who were actually there the whole period.
2. New customers leave more often. In many companies, churn is highest in the first months. When the share of new customers rises, the overall churn percentage rises too – even if no customer group behaves worse than before.
3. Large and small customers count the same. Customer churn counts every customer as one. As the company grows, the customer mix often changes too, and then customer churn and revenue churn can move in different directions.

Same month, five different answers#

Here is an imagined month in a growing company:

• 1,000 customers at the start of the month
• 200 new customers during the month
• 20 of the original customers leave
• 5 of the new customers leave in the same month
• 1,175 customers at the end of the month

Depending on what you divide by, this gives completely different churn percentages:

Hypothetical example. Customer-months assume that new customers on average arrive mid-month, and that customers who leave on average do so mid-month. Conversion to 12 months: 1 − (1 − monthly churn)¹².

Monthly churn – same month, different denominators

Show the numbers

Method	Monthly churn
End	2.13%
Average	2.3%
Start (all)	2.5%
Cohort	2%
Customer-months	2.3%

The difference between 2.0 and 2.5 percent looks small. Converted to a year, it is the difference between losing roughly every fifth and every fourth customer.

The third method is a common mistake: the numerator includes customers who left in the same month they arrived, while the denominator does not contain them. Two groups are mixed, and the figure becomes too high. The first method does the opposite: the denominator includes 200 new customers who have hardly had a chance to leave, and the figure becomes too low.

Three ways to weight correctly#

Cohort: measure the customers you had at the start#

The most robust method is to follow the customers who were there at the start of the period and count how many of them have left by the end of the period. New customers in the period are kept out completely – in both numerator and denominator.

Churn = original customers who left ÷ customers at the start of the period

This is the same logic used for GRR and NRR. The method is easy to explain, and it is not affected by how much you sell. The drawback is that churn among brand-new customers is not captured in the same period. It must therefore be followed up separately – see cohort analysis below.

Exposure: weight by time as a customer#

If you want a single figure covering all customers, including those who came and went in the same period, the denominator should be weighted by how long each customer has actually been a customer. The denominator then becomes the number of customer-months (or customer-days), not the number of customers.

Churn = all who left ÷ number of customer-months in the period

The method gives a precise churn rate per month as a customer, and it is especially useful when growth is uneven – for example when a big campaign sale lands late in a month. When growth is steady, the average of start and end gives roughly the same answer, as the example shows. The drawback is that it requires data on when each customer started and left.

Revenue: weight by what customers pay#

Customer churn says how many customers leave. Revenue churn says how much it costs. In a growing company, revenue churn is often the most important management figure, because it shows the financial effect directly.

Revenue churn = lost monthly revenue from customers at start ÷ monthly revenue from the same customers at start

Hypothetical example

2.0% / 1.2%

In the example, 2.0% of customers leave, but only 1.2% of revenue disappears – because it is the smallest customers who go.

New revenue in the period must not go into the denominator. If it does, rapid growth will make revenue churn look lower than it is – the same problem as for customer churn.

Which weighting should you choose?

1. 01

   Cohort

   Standard for management figures, GRR and NRR. Simple and robust.

2. 02

   Revenue

   When the financial loss matters most. Used together with cohort.

3. 03

   Exposure

   When you need one figure for all customers, and growth is uneven.

When new customers pull the average up#

The second mechanism is harder to see. Suppose new customers (the first three months) leave at a rate of 5 percent per month, while established customers leave at 1.5 percent. Neither group changes its behaviour. Yet the overall churn percentage rises as the share of new customers increases:

Overall monthly churn at different shares of new customers

Show the numbers

Share of new customers (%)	Overall churn
0	1.5%
10	1.85%
20	2.2%
30	2.55%
40	2.9%
50	3.25%

To management, this can look as if Customer Success has lost its grip. In reality, it is sales that has succeeded. Conversely, a period with little new business can make churn look better without anything having improved.

The solution is to split the figure by how long customers have been customers:

• Early churn (for example the first 90 days) says something about sales, onboarding and time to value.
• Churn among established customers says something about value over time, customer health and competition.

The two figures have different causes and different owners, and should be reported separately. The 90-day threshold is just an example. Set it where churn actually levels off in your own data – that may be after one month in some companies and after a year in others.

Cohort analysis: the best tool in a growth phase#

A cohort is a group of customers who started in the same period, for example the same month or quarter. Cohort analysis follows each group over time and shows what share are still customers after one, three, six and twelve months.

Hypothetical example of how a cohort table is read. The dashes mean the cohort is not yet old enough.

Cohort analysis solves several of the problems at once:

• New sales do not distort the figures, because each cohort is followed separately.
• It becomes visible whether newer cohorts are retained better or worse than older ones – in other words, whether onboarding and product are actually improving.
• The cohorts can be broken down by segment, channel or product.

The table can also be weighted by revenue: it then shows what share of the cohort's original revenue remains, and it can exceed 100% if customers expand.

Month, quarter and year#

When churn is reported for longer periods, two common mistakes occur.

Multiplying monthly churn by twelve. This exaggerates annual churn, because customers who have already left cannot leave again. The correct conversion is:

Annual churn = 1 − (1 − monthly churn)¹²

At 2 percent monthly churn, that gives 21.5 percent a year – not 24 percent.

Averaging monthly percentages. A simple average gives a small month the same weight as a large one. If one month in a quarter with 1,000 customers had 4 percent churn, and two months with 1,500 and 2,200 customers had 1 percent, the simple average is 2.0 percent. Weighted by number of customers, it is 1.64 percent (77 ÷ 4,700). In a growing company, it is the most recent months, with the most customers, that should carry the most weight.

For quarterly and annual figures, it is therefore cleanest to apply the cohort method directly to the whole period: the customers at the start of the quarter, and how many of them have left by the end of the quarter.

When the customer does not cancel#

Everything above assumes that it is possible to say when a customer has left. With agreements that renew on a fixed date, that is simple: the customer cancels or does not renew. Where the customer pays for usage or per transaction, there is no cancellation to count. The customer simply stops using the service – so-called silent churn.

The company then has to define churn itself, usually as a period without activity. Choosing the period is a trade-off, and what follows is a professional judgement, not an established standard:

• Too short a period counts customers as lost even though they are just having a quiet month. Seasonal variation makes this worse.
• Too long a period means churn is reported long after it happened, when it is too late to do anything.
• The period should follow the customers' normal rhythm. A customer who normally uses the service daily and one who uses it once a quarter cannot be measured with the same deadline.

Write the definition down, and keep it fixed over time. Report customers who come back after being counted as lost separately – otherwise both churn and new sales will be too high. Also track falling usage among customers who are still active. It is often the earliest sign. Read more about the difference between contractual and continuous renewal in What is Customer Success?

A reporting setup#

How to measure churn in a growing company

1. 01

   Lock the denominator

   Customers and revenue at the start of the period. New sales kept out.

2. 02

   Measure both customers and revenue

   Customer churn and revenue churn, side by side.

3. 03

   Split by age

   Early churn and churn among established customers separately.

4. 04

   Follow the cohorts

   Are newer cohorts retained better than older ones?

In practice, this means a monthly dashboard for a growing company should show:

1. Customer churn and revenue churn, calculated on the customers at the start of the period
2. GRR and NRR for the last twelve months
3. Early churn separately from churn among established customers
4. A cohort table, preferably weighted by revenue
5. Which method has been used – so that the figures can be compared over time

The last point is easy to forget. Churn figures from different companies, or from the same company using different methods, cannot be compared without knowing how they were calculated.

If you want to see what churn means for revenue and LTV with your own numbers, you can use the churn calculator.