Churn is not binary: The customer can stay while the revenue disappears

Short answer

Why is it not enough to measure how many customers leave?

Because a customer can stay and still take most of the revenue with them. Logo churn counts customers; revenue churn and GRR measure the value that disappears – including from the customers who stay. In transaction-based models, gradual revenue decline is often the biggest risk, but it must be investigated before it is interpreted: much of it is caused by the customer's own business, not dissatisfaction.

Most churn reports ask one question: is the customer still a customer? The answer is yes or no, and the sum becomes a percentage. That is useful, but it says little about how much value the customer base actually holds. A customer who has gone from NOK 100,000 to NOK 20,000 in annual revenue is still a customer. The company has kept the customer and lost 80 percent of the revenue.

This article explains how revenue development and logo churn fit together, how gradual revenue decline can be detected, and where the line goes for what Customer Success can and should be held responsible for. The examples are hypothetical. The observations about transaction-based customer relationships are based on generalised experience, not on a survey.

What binary churn does not tell you#

Imagine a company with 1,000 customers at the start of the year. During the year, 20 cancel. Logo churn is 20 ÷ 1,000 = 2 percent. On its own, the customer base looks stable.

Now suppose that 100 of the remaining customers have, at the same time, halved their revenue contribution. To keep the arithmetic simple, we assume all customers were the same size to begin with, with NOK 10,000 in annual revenue each.

Same year, two answers

Show the numbers

Measure	Share
Logo churn	2%
Revenue lost	7%

The churn report shows 2 percent. Revenue from the same customers has fallen by 7 percent, and over 70 percent of the loss (NOK 500,000 of NOK 700,000) comes from customers still listed as active. One figure says something about loyalty, the other about economics. They have to be read together.

Logo churn, revenue churn, GRR and NRR#

The terms are used in different ways, so it pays to be precise. Below are the definitions as used on CustomerSuccess.no. All are measured on a fixed cohort: the customers and the revenue that existed at the start of the period. New customers in the period are excluded.

GRR = 100% − gross revenue churn. NRR = 100% − net revenue churn. Whether declining usage is included in gross revenue churn depends on business model and definition – it should be stated in the report.

The terms are also explained in the glossary: customer churn, revenue churn, GRR and NRR.

A worked example with GRR and NRR#

A company has 100 existing customers with total annual revenue of NOK 10,000,000 at the start of the period. During the year:

• NOK 300,000 disappears with customers who cancel
• other existing customers reduce their revenue by NOK 700,000
• other existing customers increase their revenue by NOK 1,500,000

From revenue at start to GRR and NRR

Show the numbers

Item	Amount
Start	NOK 10,000,000
Cancellations	NOK -300,000
Decline	NOK -700,000
GRR 90%	NOK 9,000,000
Expansion	+NOK 1,500,000
NRR 105%	NOK 10,500,000

NRR is above 100 percent even though the company has lost customers and lost NOK 1,000,000 from part of the customer base. That happens because expansion at some customers is larger than the loss at others. NRR should therefore never be reported on its own: an NRR of 105 percent may mean that almost every customer is growing a little, or that five customers are growing a lot while twenty are on their way out.

The example assumes that revenue at the start and end is measured the same way for the same customer cohort. That is easy to break. Subscription businesses often use ARR (annual recurring revenue): the contracted annual value of the subscriptions. Where the customer pays per transaction, there is no contracted recurring revenue, and ARR is a misleading term. Revenue then has to be normalised, for example as actual revenue over a twelve-month period, or as the average of the last three months multiplied by twelve (a run-rate). Both can be used, but they must not be mixed: if you compare last year's actual revenue with a December run-rate, you are partly measuring seasonality, not churn.

Why this matters especially with transaction-based pricing#

In a subscription model, revenue from a customer changes in steps: at cancellation, downgrade or renewal. In a model where the customer pays per transaction, order, user action or unit of consumption, it changes every month, without anyone making a decision.

Imagine a customer who normally has 10,000 transactions a month. Over three months, volume falls to 7,000, then 4,000 and finally 2,000. The agreement is active, and the customer is listed as active in every customer report. Activity has fallen by (10,000 − 2,000) ÷ 10,000 = 80 percent.

Transactions per month for one customer

Show the numbers

Month	Transactions
0	10,000
1	7,000
2	4,000
3	2,000

The curve alone does not say why. Possible explanations are:

• The customer has lost customers of their own.
• The customer has moved volume to a competitor.
• The customer has technical problems.
• The customer has changed business model.
• The customer has a seasonal peak behind them.
• The customer is in the process of winding down the business.

Some of the explanations require action from the vendor, others require none, and one only requires the customer relationship to be recorded correctly. That is why Customer Success should investigate the cause before anyone draws a conclusion. Where activity falls all the way to zero, the issue is the same as in Silent Quitters: The customers who disappear without cancelling.

Revenue contraction: churn that happens gradually#

Revenue contraction is a decline in revenue from customers who are still customers. The term is used both for subscription downgrades and for reduced consumption in usage-based models. It is often an earlier signal than the cancellation: many customers who leave completely have used the service less for a long time first.

It is useful to distinguish between five situations:

Partial churn is not a standardised accounting or professional category. It is used here as a practical term for a customer who has moved or stopped part of their usage.

The commercial point is that the revenue is gone long before any cancellation. In the example above, the vendor has lost most of the revenue from the customer in three months, even though the customer never cancels.

How to detect revenue contraction#

Step 1Establish a historical baseline for each customer#

The baseline is the customer's normal revenue level, for example the average of the last six or twelve months. Where seasonality matters a lot, the baseline should be the same period last year.

Step 2Segment the customers#

Group customers by size, business model and activity pattern. A large customer with steady volume and a small seasonal customer cannot be assessed with the same rule.

Step 3Compare the development over time#

Compare the latest period with the baseline. Use periods of several weeks or months, so that one weak week does not trigger action.

Step 4Identify significant negative deviations#

Use thresholds adapted to the segment. The table below is an example of how different levels can trigger different follow-up.

The thresholds are illustrative, not recommendations. They must be adapted to the actual variation in the customer base – for example by looking at how large the fluctuations were for customers who later proved stable. Smaller customers often fluctuate more and need higher thresholds.

Step 5Investigate possible explanations#

Look for technical errors, cases with Customer Support, price changes, seasonality and changes in the customer's own business before contacting the customer.

Step 6Prioritise by risk and financial importance#

A 60 percent decline at a customer with NOK 1,000,000 in annual revenue is a different matter from the same decline at a customer with NOK 10,000. Prioritise by kroner at risk, not just by percentage.

Step 7Document cause, action and result#

Record what was found, what was done, and how revenue developed afterwards. That is the basis for learning which signals matter.

Not all revenue decline is churn#

Revenue from a customer can fall for at least six different reasons, and they have little to do with each other:

• The customer's own business development. The customer sells less and therefore has fewer transactions. The product works as before.
• Competitive risk. The customer moves part of their activity to another vendor, often without saying so.
• Product friction. Technical or operational problems make the customer use the service less.
• Seasonal variation. Activity follows a predictable pattern through the year.
• Price changes. The vendor's revenue changes, without the customer's usage necessarily doing so.
• Changed product mix. The customer uses other services or payment methods that give the vendor different revenue or margin.

This means Customer Success should not automatically be held responsible for all negative revenue development. The function should be able to detect and investigate the risk without owning all of its causes. A customer losing market share in their own industry is not a Customer Success problem – but it is useful for the company to know.

Customer Success should own the question of why revenue is falling – not all the answers.

Separate revenue loss from preventable revenue loss#

A useful principle is that not all revenue contraction is preventable. A customer who loses half of their own turnover will generate less revenue even if the product works perfectly. A customer who moves volume to a competitor because problems are not being solved is a different kind of risk.

A practical categorisation of causes:

Editorial categorisation. One customer can have several causes at once; record the most important.

The categorisation leads to better priorities: product-related loss is an argument in the product roadmap, commercial loss is a matter for pricing, and loss in the customer's own business says something about exposure to particular industries. It also makes it possible to measure Customer Success on what the function can actually influence, while management sees the whole revenue picture. Both are needed. A team measured only on preventable loss may be tempted to classify too much as non-preventable – which is why the “unknown” category should be tracked separately.

From binary churn to four customer states#

A simple extension of the binary model is to give each customer one of four states:

Four customer states

1. Growing

   The customer increases their revenue contribution significantly.

2. Stable

   The customer maintains the expected revenue level.

3. Contracting

   The customer shows a significant negative development.

4. Churned

   The customer has ended the relationship or meets the definition of churn.

The state can be calculated automatically from baseline and thresholds, and stored as a field in the CRM. That brings two benefits. Reporting can show how much revenue sits in each state, not just how many customers. And work lists can be built directly from the customers who have moved from Stable to Contracting. The states must take seasonality and expected activity into account, otherwise seasonal customers will move between Contracting and Growing every year without anything having changed. Revenue development can also be one of the signals in a health score, but should then be weighted together with usage and context.

How to measure the effect of Customer Success#

The number of saved customers is an insufficient measure when churn is not binary. Imagine a customer with NOK 500,000 in annual revenue who is considering moving half of their activity to a competitor because of persistent problems. Customer Success spots it, coordinates the necessary improvements, and the customer maintains their activity. In a traditional churn report nothing happens: the customer was never about to cancel.

How should the effort be assessed? It helps to distinguish between four quantities:

Hypothetical example.

It is tempting to report NOK 500,000 as saved. That is not correct. The whole revenue was never at risk, and even the NOK 250,000 assumes the customer would have moved the volume. Perhaps the customer would have stayed anyway. The effect of prevention is always an estimate, and a credible estimate requires a basis for comparison – for example customers with similar signals who did not receive follow-up. See The churn you never see for how that can be done.

A practical dashboard#

A dashboard that takes into account that churn is not binary might contain:

The first five measures are strategic: they are reported monthly or quarterly and show how the value of the customer base is developing. The last five are operational: they determine which customers the team looks at this week, and whether the work gets done. Both levels should show cohorts and segments separately. An average for the whole customer base hides exactly the differences this article is about. A dashboard used in customer meetings should also show the customer's own revenue development, not just activities.

The contract does not tell the whole story#

A customer is not necessarily retained in a commercial sense just because the contract is still active. Companies that only measure binary churn risk overlooking large changes in the economic value of their customer base – especially where revenue follows the customer's usage.

Customer Success should therefore help detect, understand and handle revenue decline before it possibly turns into full churn. That does not mean owning all decline, but making sure it is seen, explained and followed up.