Why silent customers are the most expensive

Amsterdam-based Churned builds software that learns to read this silence. The platform analyses behavioural data from thousands of customers and predicts who is about to leave, not based on gut feeling, but on patterns no human would ever spot.

But Churned goes beyond prediction alone. The AI also determines directly what the best action is for that specific customer to prevent churn, so teams not only know who is at risk, but also what they need to do today to change that.

Churned helps subscription businesses such as energy suppliers (including Budget Thuis and Vandebron), gyms such as Trainmore, and media companies and publishers such as Financieel Dagblad and RTL. 

Two brothers, but different enough to build together

Maarten and Michiel Doornenbal grew up in a small village ten minutes from Amsterdam. As children, they did not get along particularly well.

"When we were young, we didn't expect to have a company together. I wasn't very nice to Michiel when we were little."
Maarten Doornenbal [1]

That changed later. They both went to university, both ended up living in Amsterdam, but in separate homes. The distance helped. And once the bond was restored, something else became apparent: they complemented each other. Maarten is the extrovert, the commercial side. Michiel is the data scientist, quieter, more in the background. They do not get in each other's way.

"It's good that we're completely different. When brothers start a company, you don't want to be too much in each other's space."
Michiel Doornenbal [1]

The idea for Churned came from Michiel's master's thesis. During his data science studies, he investigated whether machine learning could predict which customers were going to cancel. The answer was yes. His professor Francisco Blasques saw potential. In 2020, they founded the company together with Maarten.

The pattern nobody expects

In many subscription businesses, churn only becomes visible at renewal or cancellation, and by then it is essentially too late.

There are teams that do try to look ahead. But this often happens through home-built, gut-based health scores, if they do anything at all. And where predictive models do exist, they tend to come from data science teams in terms that are difficult to translate into the day-to-day work of marketing and customer teams.

Churned bridges exactly that gap. It is a predictive retention platform that not only forecasts who is at risk, but also indicates the best action for that specific customer to prevent churn. This turns complex intelligence into something immediately practical and actionable.

Machine learning operates differently from classical analysis. It searches for correlations a human would never think of. And sometimes those correlations are counter-intuitive.

"It's not always the case that more product usage is better. What we often see is that usage actually goes up just before the moment of churn. That's unexpected behaviour."
Michiel Doornenbal [1]

A customer who suddenly logs in far more frequently than usual may not be enthusiastically engaged. They may be exporting their data. They may be looking for a way to downgrade their subscription. The pattern is the same; the meaning is different. Churned trains models to see that distinction.
In B2C companies, this translates into behaviours such as comparing tariffs, adjusting contract options, or viewing cancellation pages more frequently.

The market where people are too expensive

In September 2024, Churned raised €2.5 million in a round led by Newion and Volta Ventures. [2] The funds are earmarked for international growth and further development of the platform.

The strategic focus is clear. Churned targets what it calls high-volume, low-margin subscription markets, think streaming services, publishers, e-commerce with recurring payments. In these markets, margins are thin and human intervention per individual customer is simply too costly.

Patrick Polak of Newion notes that Churned's technology replaces rule-based systems with AI, specifically aimed at companies with large customer bases where every percentage point of churn matters. [2]

The arithmetic is straightforward. A one-percent reduction in churn can increase a company's profitability by five to fifteen percent. [3] That may sound abstract, but for a company with tens of thousands of subscribers, it is the difference between growth and stagnation.

At Uptrends, a Dutch SaaS customer of Churned, the platform led to twenty percent less annual churn and a 74 percent decline in revenue loss from cancellations. [4]

The human element remains essential

Maarten does not see AI as a replacement for the customer success manager, rather as a collaboration between different kinds of intelligence. A predictive model that identifies which customers are at risk. A prescriptive model that advises which action to take. And, in time, generative AI that executes the action.

But the strategy, the playbooks, the relationship with the customer, that remains human work. The AI handles the heavy computational lifting; the human sets the direction.

"It still needs the human to create the playbooks and come up with new actions. The AI model can then look at everything you've done and see what worked best."
Michiel Doornenbal [1]

The unglamorous problem as a goldmine

While the world is fixated on generative AI and the latest chatbots, Churned is building something that rarely makes the front pages. Predicting customer churn does not sound spectacular. Nobody posts about it on LinkedIn. But it is a problem that costs the subscription industry billions of euros each year.

That holds a lesson for the Dutch startup scene. The most defensible positions are often found not in the hype, but in the infrastructure beneath it, in the specific, measurable problems that keep large companies up at night. Churned is not competing with OpenAI. They are competing with spreadsheets and gut instinct. And that is a race you can win.

Two brothers who could not stand each other as children. A professor who read a master's thesis and thought: there is a company here. An insight that silence does not mean satisfaction. Five years later, that idea has raised millions and is expanding into Scandinavia and the United Kingdom.