Request a potential analysis
EN
Menu

Churn prediction: spot churn before it happens.

Predicting, measuring and acting on customer churn in retail in good time.

Good customers rarely leave with a loud goodbye. They come less often and eventually not at all. Vittorio, the sales assistant powered by artificial intelligence (AI), identifies good customers who have not been in for a long time and suggests a personal message to the responsible employee while it can still make a difference.

Official WhatsApp Business Platform Hosted in Germany Developed in Hamburg
Good morning, Tobias.Tuesday, 10:30

3 contact suggestions for today

Mr Krämer Not seen for a while Used to come every month, not for four months now
Hello Mr Krämer, from Friday our game menu is back, the one you enjoyed so much last autumn. May I keep your table by the window for you?
EditLaterSend
Mrs AdlerTable freeOn the waiting list for Saturday evening
The Schmitt familyAnniversaryCelebrated their wedding anniversary with us last year, as stated by the guest

Key points at a glance

  • Churn means customers leaving. Churn prediction is forecasting which customers are likely to leave.
  • The churn rate is the share of customers lost in a period, relative to the customer base at the start.
  • In retail, there is no cancellation. That is why every business has to define when a customer counts as churned.
  • The most important warning sign is a gap since the last visit that is clearly longer than the customer’s own rhythm.
  • A prediction is only useful if someone acts on it: with a personal message from the familiar contact person.

What does churn mean and what is churn prediction?

Churn is the term for customers leaving a business. Churn prediction means forecasting which customers are likely to leave in the near future. The aim is to identify these customers while they can still be retained. Dealing with the results is called churn management.

The topic is well known from contract-based industries such as telecommunications, insurance or subscriptions. There, churn is a clear event: the cancellation. In retail, hotels and restaurants, this event is missing. Customers leave silently. That makes prediction harder and at the same time more important.

Calculating the churn rate

The churn rate measures how many customers are lost in a period:

Churn rate = customers lost in the period ÷ customers at the start of the period × 100

An illustrative example: at the start of the year, a fashion house has 2,000 regular customers. By the end of the year, 260 of them have not bought again, although they used to come several times a year. The churn rate is 260 ÷ 2,000 × 100 = 13 per cent.

The calculation is simple. The definition is the hard part: when does a customer count as churned in retail? A fixed deadline for everyone rarely fits. A rule that takes each customer’s rhythm into account makes more sense, for example: no purchase for more than twice their usual interval. There is no universally good churn rate; it depends heavily on industry and buying rhythm.

Churn rate in retail: an example with buying rhythm

An illustrative example shows how strongly the definition shapes the result. A jeweller has two regular customers who have both not been in the store for 14 months:

  • Mrs A. has bought roughly every four months for years. For her, 14 months without a visit is more than three times her usual interval. She has very probably churned or is about to.
  • Mr B. buys every two years for a fixed occasion, most recently a watch for a milestone birthday. For him, 14 months is normal.

With a fixed deadline of twelve months, both would count as lost. A rule based on each customer’s own rhythm, on the other hand, only flags Mrs A. as at risk. She is exactly the one who should receive a personal message today.

Key figures around the churn rate

The churn rate alone tells only half the story. It becomes meaningful with these key figures:

  • Retention rate: the share of customers who stayed during the period. It is the counterpart of the churn rate: 13 per cent churn corresponds to 87 per cent retention.
  • Revenue churn: the share of revenue lost with the churned customers. If a business loses few but especially good customers, it is considerably higher than the churn rate.
  • Reactivation rate: the share of at-risk customers who buy again after being contacted. It shows whether the countermeasures work.

Churn rate and customer lifetime value

Every churned customer shortens their relationship with the business and thus their customer lifetime value (CLV). A lower churn rate therefore has a double effect: it keeps revenue in the business and increases the value of each regular customer over the years. Contacting customers who have a high value and are drifting away pays off most.

How churn prediction works

There are three common approaches, from simple to complex:

  1. Rules. Fixed thresholds such as “no purchase for 12 months” or classes from RFM analysis. Quick to implement, but crude.
  2. Statistical models. They estimate from past buying behaviour how likely a customer is to still be active. This requires clean data over several years.
  3. Machine learning. Models learn from many features which customers churned in the past and apply this to today’s customers. Powerful, but hardly manageable without specialists.

For a business with regular customers, the method matters less than what happens next. A list of probabilities does not bring a single customer back.

From prediction to action

An at-risk customer needs a reason to come back and a person who gets in touch. Ideally the contact person they know, with an occasion that fits their history with the business. In large systems, the question of the right next step is called next best action.

What such an approach can look like, and what to avoid, is set out on the customer reactivation page.

What Vittorio does with it

Vittorio identifies good customers in your sales data who have not been in for a long time, as well as other occasions such as services due or recurring appointments. Instead of a list, the responsible employee receives a suggestion every morning with the reason and a ready-made draft. They decide whether and how to write. The free potential analysis with your own data shows how many good customers have not been in touch with you for a long time.

Example scenes. All names of people and businesses are fictitious.

Churn with and without contracts.

Why the familiar recipes from telecommunications and subscriptions only partly fit retail.

FeatureContract-based businessRetail, hotels, restaurants
How churn becomes visiblethrough the cancellationnot at all, the customer silently stays away
When the company noticesimmediatelyoften only after months, or never
What is measuredcancellations in the periodtime since last purchase relative to the customer’s own rhythm
Typical responseretention offer on cancellationpersonal message with an occasion, before the customer is gone for good

How to spot churn early.

The gap gets longer

The last visit is clearly further back than the customer’s usual rhythm. In retail, this is the strongest signal.

Visits become less frequent

Someone who used to come every month and now only every quarter is drifting away, even if they have not gone completely.

Spending falls

The customer still buys, but less per visit. Often they now buy the rest elsewhere.

The service happens elsewhere

The inspection, the maintenance or the watch check does not happen, although it is due.

The contact person has left

When a long-serving employee leaves the business, their customers are particularly at risk. The use case the new contact person shows an example.

A complaint or a disappointing experience can also be a trigger. It rarely appears in the sales data, but your team often knows about it.

Read on.

Resources

RFM analysis

Ranking customers by last purchase, frequency and revenue, with an example.

Go to RFM analysis

Discretion comes first.

What your clients entrust to you stays in your house.

People decide

Every message your house sends on its own initiative is approved by a member of staff. Vittorio never writes on its own authority.

Consent only

Contact exclusively with customers who have expressly agreed to WhatsApp.

Official platform

Messages are sent via Meta’s WhatsApp Business Platform, not via private accounts.

Hosted in Germany

Designed for use under the General Data Protection Regulation (GDPR), with a data processing agreement.

First, see what your existing customers are worth.

Before you decide on any software, we analyse your existing sales data, pseudonymised and contractually secured. After ten working days, we present the results to you in person.

  • How many good customers have not been in touch for a long time
  • Which occasions are coming up in the next 90 days
  • How much revenue potential lies in them
Example of a potential analysis

Jeweller, two stores

1,240Customers with purchases over €5,000418 of them not in store for more than 18 months
96Watch checks due in the next 90 daysPurchases 10 to 14 months ago, with consent
37Occasions from customer detailsAnniversaries, birthdays, wedding rings

Illustrative figures. The analysis shows your own numbers, based on your data.

Frequently asked questions

What does churn mean?

Churn is the term for customers leaving, meaning the loss of customers within a period.

What is churn prediction?

Forecasting which customers are likely to leave in the near future. The aim is to identify these customers in good time and contact them personally.

How do you calculate the churn rate?

Customers lost in the period divided by the customers at the start of the period, times 100. In retail, you first need to define when a customer counts as churned.

What is a good churn rate?

There is no universal figure. It depends on industry, buying rhythm and the definition of churn. How your own churn rate develops over time is more meaningful.

What data do you need for churn prediction?

At minimum customer, purchase date and amount over a longer period. Service and appointment data and the responsible contact person are also helpful.

What is churn management?

Dealing with at-risk customers: identifying, prioritising, contacting and measuring the result. In retail, this mainly means contacting good customers personally before they start buying elsewhere for good.

What does a churn rate of 10 per cent mean?

Of 100 customers who were active at the start of the period, 10 had churned by the end. Whether that is a lot depends on industry, buying rhythm and your own definition of churn.

How are churn rate and customer lifetime value related?

The more customers churn, the shorter the average customer relationship and the lower the customer lifetime value. Lowering the churn rate increases the value of your regular customers.

Let’s talk about your existing customers.

30 minutes with one of the founders, in person at the H1 in Bielefeld, by video or directly in a Teams chat.