- The classic churn rate formula assumes a churn probability that is constant over time, which is very often wrong.
- A declining company can show an improving churn rate, and a hypergrowth company a churn rate that stays flat, without customer lifetimes having changed.
- The adjusted rate divides churned customers by an average of the number of customers over the period, but its results change depending on the time window chosen.
- Tracking by acquisition cohort neutralizes the effect of acquisition, at the cost of many more numbers to track.
- To manage retention: NPS, repeat purchase rate, revenue per cohort and CAC payback period.
Churn rate is a fundamental metric for every business, and even more so in B2B, where average deal sizes are usually higher. But it is often miscalculated, and misread.
The traditional formula we get served everywhere, the number of customers lost over a given period, leaves out a lot of important factors. Worse, it can be misleading.
How do you calculate churn rate? How do you optimize it? Which KPIs should you track to improve your retention rate? We cover it all in this article.
Calculating churn rate: the classic formula

Retention rate VS churn rate
Customer retention rate lets a company measure customer retention, meaning the percentage of customers it has kept over the long term. The idea is to be able to optimize customer loyalty.
To calculate it, we suggest our very thorough guide to customer retention.
The CRR is often set against the churn rate, or attrition rate, which is the percentage of customers a company has lost over a given period.
The question of the retention rate unit
But in practice, customer retention is also measured with the churn rate. There is no single true unit for retention rate, and you can calculate your customer retention with these different metrics:
- The customer retention rate, whose formula we walk you through in our guide to customer retention. The retention rate is always between 0 and 100%.
- The license retention rate, which can be negative, even for a company that is doing well.
- Finally, and this is where we get to the heart of the matter, the churn rate, which unlike the CRR can be negative.
Go furtherWe can only recommend our complete guide to customer retention.
The problem with the classic way of calculating churn rate
It is a bit complex, but it is worth taking apart the classic way of calculating churn rate, which is not always relevant because it starts from a false assumption and will steer your efforts toward the wrong solutions.
A fundamentally false assumption …
The churn rate calculation assumes that a customer always has the same probability of leaving at any moment, however long they have been subscribed. That is very often wrong. In fact, a churn probability that stays constant over time implies that you will have no customers left in the long run.
If a customer’s churn probability is constant over time, it means customer lifetimes are modeled by a geometric distribution.
If you have a constant churn rate of c per month, the probability that a customer stays subscribed for n months is (1-c)^n: that is the geometric distribution. If customers can leave the subscription at any moment, we are in continuous time and need to use the continuous-time analogue: the exponential distribution.
What you need to remember is that it is very unlikely that someone who has just become a customer has the same probability of canceling as a customer who has been with you for 3 years. Thankfully, otherwise your customers would have very short lifetimes.
The classic way of calculating churn rate completely ignores this fact and assumes a probability that is constant over time, which can lead to fairly serious errors.
… that produces misleading numbers
Now we need to understand why correct modeling matters: to do that, we take two examples, a company that is struggling and a company that is growing fast, to see what their churn rates say when they are calculated badly.
Keep in mind that in each of the examples below, we simulate lifetimes from the same distribution of customer lifetimes, and that this distribution does not change over time.
Example: a company that is doing very badly, where the number of new sign-ups per day drops by one every day. How will this show up in the churn rate?
With business shrinking, the churn rate seems to improve because subscriptions are fewer and fewer, and older and older.
This is clearly a company going through a rough patch, but if you look at the churn rate, it looks excellent! The daily churn rate keeps falling, even though we know customer lifetimes do not change in our model.
This sharp drop in the churn rate is simply the consequence of the company not getting new customers.
On top of that, since it is not growing, a large share of its customers have been around for a long time and are therefore less likely to churn, which means our churn rate falls more than it otherwise would.
This change at the population level happens even though there is no change at all in the underlying individual customer lifetimes.
Second example: the scenario of a company going through a hypergrowth phase. Its customer lifetimes stay the same, but the number of new sign-ups per day grows faster than linearly.
With business growing, the churn rate seems not to change, only because most subscriptions are new.
Even though customer lifetimes stay unchanged, the churn rate chart is flat. An investor would frown and say the company is doing nothing to improve its retention rate.
In reality, the only reason the chart looks “bad” has nothing to do with churn: it is because we are doing incredibly well at getting new sign-ups.
Relying on this metric for the company suggests that the retention rate is stable, or even declining, simply because it is polluted by the uneven distribution of customer acquisition over time.
In shortThe key takeaway is that this way of calculating only lets you estimate the average age of your customers, which says nothing at all about a company’s health if the distribution of customer acquisition dates is not constant. Growth or decline in the volume of customers acquired over one period affects the measured customer lifetime in the next period, when those two things should be independent.
Go furtherIf you want a quick refresher on the different metrics to calculate to optimize your sales, we suggest:
Adjusted methods for calculating retention rate
Ideally, you would run real statistics to determine the probability, at any given moment, that each of your customers leaves you for a competitor.
In real life, unless you have a huge number of customers, that is impossible. And even with a huge number of customers, you will not be able to segment – which is fundamental to optimizing your retention rates and making the calculation useful.
So in practice we use calculation methods that help limit the margin of error:
1. The adjusted churn rate
To adjust the churn rate and so account for significant monthly growth, you can start by taking the midpoint of the number of customers for the month, rather than using its value on the 1st of the month.
Here, you divide the number of churned customers by an adjusted average of the number of customers over the whole period studied, to smooth out the effect of acquisition over the time of the study.
Pros and cons:
- This approach deals with the growth problem by normalizing changes in the total number of customers over the time period. You now have a more stable platform on which to base your churn rate, since the time window for the total number of customers is the same as the one for churn.
- However, while this approach to calculating churn rate deals with the growth problem, it fails to adapt to different time windows. Using the same calculation and the same data, you will get very different answers for a daily, weekly, monthly and quarterly churn rate.
Example: take, for instance, the numbers of an imaginary B2B SaaS company like the ones in the table below.
Even with a different number of customers at the start of each month, we get the following churn rates:
- August: 625 / 12,187.5 = 0.0513
- September: 844 / 16,453 = 0.0513
- October: 1052 / 20,505 = 0.0513
- Over the quarter: 2521 / 16,239.5 = 0.1552
The main problem with this approach to calculating churn rate is that it makes assumptions about the data. If you calculate it over 3 months, you get a churn rate of 15.52%. Divide that figure across the 3 months, and you get 5.17%, which is very close to the individual monthly customer churn rates. So far, so good.
But what happens if you do not have exactly the same numbers for every month? Let’s make August a bad month for our imaginary B2B SaaS company. This time, it only gets 100 new customers, 2 of whom churn.
The behavior is the same in terms of churn (5% of existing customers and ~2.5% of new customers), and when calculated individually, each month shows almost the same churn rate: 5.12% in August, 5.13% after that.
But when it is calculated by quarter, you get a 3-month churn rate of 13.72%, which, divided across each month, comes to 4.57%.
- August: 502 / 9,799 = 0.0512
- September: 605 / 11,795.5 = 0.0513
- October: 825 / 16,080.5 = 0.0513
- Quarter: 1932 / 14,084 = 0.1372
Now the monthly churn rates no longer match the quarterly churn rate, even though they use exactly the same data.
In shortChurn is never all that useful. A good churn rate should stretch or shrink with the length of the period it measures, while still giving comparable results.
2. Calculating churn by cohort

There are two types of cohorts: behavioral cohorts and acquisition cohorts.
Behavioral cohorts are grouped by specific actions that users take in your product. They let you define a specific behavior of course, but also a period of time (during which that behavior applies).
They are useful for understanding how users behave with your product, but to analyze churn rate, what you care about first and foremost are acquisition cohorts.
Acquisition cohorts are based on when a user signs up for your product (quite simply, people who signed up in February, for example, belong to the same cohort).
Acquisition cohorts can help you determine customer LTV, average time to churn and other retention-related information.
To avoid acquisition bias in your churn measurement, you can simply hold acquisition fixed and study cohorts by acquisition period to see how they evolve over time.
Every month, you look at all the cohorts acquired in M-1, M-2, M-3, … up to M-12 in general.
Pros and cons:
- The big advantage is that it gives you a relatively simple way to calculate churn rate every month. And this time, the calculation method presented at the start actually works.
- The big drawback is that it means a lot of numbers to track instead of a single magic KPI.
It is also closer to operational reality, because it is pretty hard to measure the impact of small actions on one big global KPI. With cohorts, you get a feel fairly quickly for what has an impact on retention.
Watch outEach cohort can be very different for all sorts of reasons. You will always be tempted to attribute it to actions you have taken (changes to the offer, the price, etc.), but be careful to also think about external factors. For example: a competitor over- or under-investing in a given month, seasonality, the news, …
Go furtherWant to manage your customer relationships better?
- Everything you need to know about CRM
- Why use a sales CRM?
- The detailed comparison of the best CRM software.
The KPIs to track to improve your retention rate

1. The NPS score
Net Promoter Score is an indicator of overall satisfaction and loyalty to a brand. By calculating your overall NPS, you know whether your customers are satisfied with your products or services and ready to recommend them to others.
Created by Bain & Co, NPS has gone beyond being a simple metric and is now used as a company-wide system to drive customer focus. It lets you estimate potential growth through customer loyalty and referrals by comparing your score with your revenue growth rate and your customer churn rate.
Unlike most metrics, NPS is calculated from a question that is common in customer surveys: “How likely is it that you would recommend our company to a friend or colleague?”
Customers are asked to rate how likely they are to recommend your company to their family and friends on a scale of 0 to 10. Customers are split into three categories based on the scores they give:
- Promoters: 9-10
- Passives: 7-8
- Detractors: 0-6
Once the customer scores are recorded, use this formula to determine your NPS:
Net Promoter Score = %Promoters – %Detractors
2. The repeat rate
The repeat rate, or repeat purchase rate, is the percentage of customers who repeatedly buy other products and services from your company, which makes it a good indicator of customer loyalty.
One useful application of this metric is capturing data for different demographic categories of customers. The results can help you refine your customer personas and improve your marketing to those target audiences.
The repeat rate can be calculated for any period, which makes it useful for looking at small time windows or at the big picture. To determine your repeat purchase rate, use this formula:
Repeat purchase ratio = Number of returning customers / Total number of customers
3. Revenue generated by each (acquisition) cohort
Knowing the revenue generated by each acquisition cohort is essential for your business.
- If it is going up, it shows that your teams and your campaigns are succeeding and that your customers get significant value from your company.
- If it is flat, it means you need to put more money and effort into customer retention and account-based marketing.
- If the number is going down, your company may be in trouble – so you need to find out what is wrong and fix it, fast.
To calculate the revenue growth rate of your existing customers, you need to know the monthly recurring revenue (MRR) of your existing customers, without including new sales. The formula is:
Existing customer revenue growth rate = (MRR at the end of the month – MRR at the start of the month) / MRR at the start of the month.
4. “Payback”
Finally, you can calculate CAC payback and debt. Debt takes many forms within a company. It can be a bond or a traditional long-term bank loan. In general, there is a principal balance that you repay over time, plus some kind of interest.
But you can also think of customer acquisition costs (CAC) as another form of business debt, with one small twist: you have a balance or principal amount, the amount spent to acquire a new customer, and an interest component. In this case, the interest is the opportunity cost of the money tied up in CAC that could have been spent elsewhere (for example, on a new product).
It is a metric people pay little attention to, even though it shapes a company enormously in a high-growth context.
To calculate it in a meaningful way, you need to project a cohort’s existing revenue, which can be a problem since it assumes a known churn rate. Except you do not know it, precisely because you are working on it (you are trying to bring it down).
In practice, you can use two or three approximations depending on your churn curve. In SaaS, for example:
- Churn is usually very high at the start of the customer relationship (onboarding).
- It then settles much lower (thankfully), around 3-5%, from the 3rd or 4th month onward.
- The other key point is usually the 13th month, when it shoots up because of annual non-renewals and a noticeably different buying cycle (churn on those subscriptions being, in fact, zero the rest of the time).
To project each cohort’s revenue into the future, you really have to take these structuring assumptions into account and retroactively validate the assumptions made for each cohort.
For example, in the March 2022 reporting, you validate the retention rate of the December 2021 cohort and of the March 2021 cohort, and adjust the assumption for the following months if needed.
Go furtherNeed advice on improving customer loyalty?
- The best customer support software
- Nurturing your customer relationships at a trade show
- Turning your customers into advocates to improve your customer acquisition cost
The questions we get over and over.
Why is the classic churn rate formula misleading?
It assumes a customer always has the same probability of leaving, however long they have been a customer. The number then depends on the average age of your customers: it falls when acquisition slows down and stays flat when it speeds up, even if customer lifetimes do not change.
How do you calculate an adjusted churn rate?
You divide the number of churned customers by an adjusted average of the number of customers over the period, for example the midpoint of the month rather than the value on the 1st of the month. This smooths out the effect of growth, but the results differ depending on whether you calculate by month or by quarter.
What is an acquisition cohort?
It is a group of customers who signed up during the same period, for example everyone who arrived in February. Tracking each cohort month after month, usually from M-1 to M-12, lets you measure churn without acquisition bias.
Which KPIs should you track to improve your retention rate?
Net Promoter Score, repeat rate (or repeat purchase rate), the revenue generated by each acquisition cohort and the time it takes to recover customer acquisition cost, the payback.
How is NPS calculated?
You ask customers whether they would recommend the company, on a scale of 0 to 10. Scores of 9 and 10 are promoters, 7 and 8 passives, 0 to 6 detractors. NPS is the percentage of promoters minus the percentage of detractors.





What the community says.