How To Predict Customer Churn: 6 Proven Methods

Article by

Danny Maguire

Customer Success Specialist @ Frankli

No items found.


September 29, 2023

“Churn” is the one word that will keep anyone in the world of customer success awake at night. Predicting and preventing churn is our #1 job.

Customer-acquisition costs in SaaS are roughly 4-5x the costs of retaining customers. So, once we’ve secured a customer, it’s imperative that we keep them.

At, this is exactly what I’ve been tasked with (We’re a performance and employee engagement platform, for anyone who’s interested). As I’ve discovered in my own work, churn is bound to occur at some stage and is a natural part of the sales cycle. Whether preventable or not, it’s always disappointing. In this article, I’ll focus on how to predict what customers may be headed towards the exits.

What causes customer churn?

Before we talk about how to predict churn, we should consider what causes churn. While there are hundreds of possible reasons given by customers on their way out, the core issue usually comes from (and there’s a clue in our titles...) a customer’s success.

Greg Daines discusses this in his book, The Effortless Experience.

Customer Success exists to ensure customers get good results with your product. You have to earn their renewal by delivering value.

Once we know exactly what our customers want to achieve, we can then tailor our approach to bring consistent value.

In my own experience, the three areas that lead to a customer failing to realize the value of a product are:

1. Bad product-market fit

In my opinion this is the most direct route to voluntary churn you’ll find, and it's a very common problem. Especially for startups. It means you’re attracting customers who are destined to fail to find value in your product. Ensuring your company has a correct product-market fit and a clear ICP for your sales team to target, will make life as a CS agent far more straightforward.

2. Failure to demonstrate ROI

An important thing to remember in SaaS is that people want the outcome, not the tool. Therefore it’s necessary to highlight the value our software delivers on a consistent basis. Otherwise, customers who are looking to tighten their belts are far more likely to cut ties with you. We can prove the value our SaaS product delivers either by direct communication with our Champions or via reporting tools within the software.

3. Poor adoption

One way to guarantee a lack of value for your customer, is if a majority of their organization fails to even make a start. Invest in your onboarding to ensure customers are as deeply engaged with your tool as possible. Another truism of Daines’ is that the customers who adapt most to your product will be the most successful.

What do you need to be equipped with to predict churn?

Data. Data is the main thing your CS team is going to need if it is to have any shot at predicting which customers may be headed towards the exit. Both quantitative and also qualitative customer data will be the fulcrum upon which you leverage your CS strategy. The exact data points you'll want to be equipped with will depend on your product & business model. Broadly though, you would ideally have:

  • Product usage data
  • Satisfaction metrics like CSAT & NPS
  • A customer health scoring model
  • Customer company information (size, type, industry etc.)
  • Engagement & advocacy metrics

Let's dig deeper into some actionable ways to predict churn.

6 ways to predict customer churn

1. Segment your customers

By segmenting our customers we can identify which types tend to churn the most. To be as exact as we can, use as many segmentation categories as possible:

  • product-customer fit
  • customer industry
  • no. of users per customer
  • contract length
  • customer journey stage
  • (etc.)

Once we’ve broken things down to a manageable level, we can begin to search for patterns among at-risk customers. This will enable us not only to step in at an early stage, but also help us decide what is the most effective solution to employ, depending on the situation.

2. Look for dips in feature usage & adoption

One of the biggest indicators that a customer is likely to churn is low product usage.

I asked James Leggett (CSM at Velaris) what he felt was key to predicting churn. He said:

Knowing your customer’s activity means an understanding of your customer themselves. It will allow you to action successful mitigation tactics.

Figure out which features are most used by your paying customers, and take an activity baseline. For example, At Frankli I monitor numbers such as the amount of users seen in key spaces during a certain period. Then, you can monitor activity, and look for substantial drops below the baseline.

When that happens, you can alert a CSM to take certain actions (it could be as simple as reaching out, or triggering an automated email) to see if they can offer proactive support.

3. Implement other ongoing data analysis

Outside of feature usage, there's many other customer success metrics & data you can monitor to help with predicting churn. This can either be done in-house by your own team, or by availing of one of the many customer success platforms on offer with churn prediction features like Gainsight, ChurnZero or Velaris. Whichever way you go about it, you’ll want to include a mix of data inputs to deliver accurate results. This data can be derived from a number of sources such as:

  • Customer survey outcomes: This involves results from NPS (Net Promoter Score) or Customer Satisfaction surveys.
  • Customer engagement: Have they been willing to provide referrals or client testimonials? Have they allowed you to use their logo?
  • Company traits of churned customers: Are companies in a particular industry, or of a particular size, significantly more likely to churn?

By compiling these key metrics, you’ll be presented with a far clearer picture of the true situation.

4. Implement customer health scoring

Once we have both segmentation and a model that is feeding us predictions based on both qualitative and quantitative data, we can then begin applying customer health scores. These are an ongoing process and while they might not be perfect at first, as you learn to tweak your model and gather more data, they will become more and more valuable.

It’s also important to highlight that we cannot grade our customers all on the same basis. We have to take into account what the customer is trying to achieve with our product and whether our data tells us if they are succeeding or not. Health scores can be a powerful visual indicator of whether it’s time to step in with a re-engagement strategy or conversely, if a customer is a prime upsell opportunity.

Once you have a reliable model, you can predict that churn is likely to occur below a certain score, and trigger certain playbooks & automations to try preventing it.

Here's a list of churn prediction software that can handle this use case.

5. Have regular 1:1 customer interactions

The first thing I wanted to do as part of our customer success approach at Frankli was to try and regularly get face to face with our customers. Mainly because I feel any issues, upsell opportunities or feedback sessions are (unsurprisingly) far more effectively conducted on a call vs an email.

Building these relationships is a key part of what we do. These conversations will give you insights into the customer’s perspective, helping you to manage expectations and develop relationships with other internal stakeholders. Gainsight highlights this with their 11th Law of Customer Success stating that “CS is human first.”

These interactions provide you with the opportunities to ask questions & get insights into their likelihood to renew or churn. Importantly, having built a relationship with stakeholders, you drastically increase the chances that they'll actually be open & honest with you about their plans & feedback.

Additionally, if customers start to miss their regular 1:1 calls, that in itself is concerning -- and may be a predictor of churn.

6. Analyze churn reasons from past customers

This one should be fairly obvious. However, we must go back to the very beginning of a customer’s journey and see where it all began to go wrong. Were they never a good fit? Did you fail to understand their needs initially? Or, perhaps you failed to demonstrate your ability to deliver results? Whatever it is that went wrong, work to make sure it doesn’t happen again. Churn is natural and sometimes it may be unavoidable, so try to turn these negatives into positives and take learnings to create better predictions and a better experience for current and future customers.

Final thoughts

So there you have it, 5 ways to predict customer churn.

Get close to your customers, understand them and help them achieve things they couldn’t without you. Take a look at your customers and if you don’t think you’re nailing those pointers, they’re most likely your key candidates to churn next. Re-engagement is always a tough task, so I’d advise getting the work done early to ensure each customer has the best possible start. This will make maintaining their success a far easier ride.

Remember: customer success is constantly evolving. Keep adjusting things and listening to whatever your data is telling you as your platform and customer base continue to change.