In my experience working at Airtel, Dell, and Freshworks, I have moved from revenue planning to business operations, to growth analytics, and then product analytics. But all along, the work stream of churn management has kept me company. And now that I am building my own SaaS startup, there is no way I can not think about churn. 😉

Churn management is a perpetual exercise: an organization needs to be always on its toes to manage churn, and can never be done with it. But it is also a rewarding exercise: keeping an eye on churn helps companies be aware of groundswells related to changing user preferences or dipping user experience. Long before market research reports or lagging indicators such as revenue growth start pointing to them.

For a SaaS company, churn is as inevitable as death or taxes. But how you choose to deal with it can make a big difference to how much it impacts your business.

Address churn proactively, before it becomes a problem—and you’ll save money, your reputation, and your relationships with your customers. Here’s my take on how to manage churn before it becomes too hot to handle.

Why do customers churn?

The reasons for churn fall broadly into three buckets.

1. Reasons beyond your control

There are certain kinds of churn you can only regret, but not fix: for example, if your customer has gone out of business or is going through budget cuts that necessitate them to downgrade or cancel your subscription. Focus your energies on the other two buckets, which are well within your capacity to fix.

2. Reasons within your control

Here are some other reasons your customers might churn, reasons that are well within your control to prevent.

• Overselling the product, thereby setting unrealistic expectations for customers, leading to disappointment when the product doesn't meet those expectations.
• Not onboarding diligently, which means you are not setting up the customer to use the product to its optimal value.
• Lack of product adoption due to inadequate product marketing, which means customers are using a limited version of the product and aren't extracting enough value.  
• A product that is not ‘sticky’ or lacks hero features that can keep users engaged and committed over the long term.
• Slowness in product development and delays in providing support, which may lead customers to go looking for more responsive options.
• Failure to keep pace with customer needs and the constantly evolving tech landscape.
• Glitches in migration to a new version of the product, leading to disruptions, data loss, or other issues that negatively impact the customer experience.

3. Controllable reasons that appear uncontrollable

These are the ones that tend to be missed because at first glance, they are attributed to causes beyond our control. But dig deeper, and you’ll realize that there are aspects we have influence over.

Here’s a common example: your product evangelist within the customer team moves on and within weeks, the customer churns. At first, it looks like the loss of your evangelist is the reason for the customer churning. But look more closely and you may realize that you have either not engaged with other client stakeholders enough, not won the trust of other leaders, or failed to demonstrate sufficient value. If you had, irrespective of who is at the client end, churn would not happen.

Customers outgrowing the product is another instance where the cause for churn is seemingly uncontrollable. But the truth is that your product is not able to cater to the evolving needs of your customers whereas a competitors’ product can.

Delays in responding to customer needs or matching competitors’ offerings is the root cause—perfectly controllable.

Of course, there are times when you decide that the product should not move in a certain direction. That’s an anticipated loss and should not cause alarm.

The qualitative solution to churn

If you look at the reasons I’ve discussed so far, one thing is clear—reducing churn cannot be the responsibility of any one team. It is a cross-functional responsibility that should not be left to the customer experience or customer support teams.

Here is a sampler of how each function can influence the overall churn rate.

• Marketing: What sort of customer segments are being brought in through marketing campaigns? Are they profitable segments?
• Sales: Which of these are deemed worth following up by the sales team and how do they pitch the product? Are they overselling it?
• Product: How do the decisions made by the product team in terms of features and functionality impact customer engagement and eventually retention?
• Customer support: How is support structured and offered in a way that truly supports the customers?

In short, managing churn should be a concerted effort of all these functions. When each of these functions focuses on the larger goal of establishing sustainable, long-term engagements, that’s when your customers stay with the product. Your churn rates come down organically.

The quantitative solution to churn

There’s one more handy tool in your churn management kit: churn prediction modeling.

You can build a logistic regression model to calculate the likelihood of churn, based on easily available data. You don’t have to look for anything too costly to implement or too difficult in terms of data collection. These factors can include customer attributes, in-app activities, and technical support tickets raised—to name a few. Beyond a point, adding more variables to the model may not bring significant incremental value.

With modeling, all of this data can be converted into a 360-degree view of your churn risk. The GTM team can use this analysis to win back at-risk customers (in case you’re wondering whether it is a tall task to woo customers back, you’d be pleasantly surprised. As long as there is honesty and persistence in the win-back process, many customers are usually ready to give you a second chance).

The prediction model—whether it is K-means, logistic regression, decision tree or any other—will gain currency with your teams when it's explainable and understandable. Pick a model which is easily explainable, in order to get the trust of the GTM team working on this initiative. Once they understand and trust the model, they will believe in the initiative. The rallying behind the initiative is as important as the efficacy of the model.

The Confusion Matrix

Presenting the prediction summary as a confusion matrix helps in gaining confidence among stakeholders.

The confusion matrix is based on probability thresholds of churn. What’s important to note is that these thresholds can change based on how many people there are to engage with at-risk customers. For instance, if you have a tiny customer success team with limited bandwidth to attend to at-risk customers, you need to focus on improving precision. i.e. minimizing false positives. But if the team is well-staffed, then you can focus on improving recall. i.e. minimizing false negatives.

To put it simply, if the cost of missing out on predicting a large customer churn is high, you can staff up and focus on recall. If the cost of staffing up is really high, focus on high precision and use segmentation to identify high value customers to target your efforts.

The True Role of Prediction Models

I have built several churn prediction models over the years. In the early days, I’d get very excited by the prowess of such models. But with experience, I’ve come to put them in their place, to realize how they should just be one part of a bigger puzzle.

Sure, you can create models to predict when churn will happen and put in a lot of effort in winning back those customers through various programs. And sure, these models can give your data team a sense of validation, the satisfaction of seeing their insights come true.

But I’d put my money on the comprehensive, team-sport approach to retaining your customers. If I have to draw an analogy, it would be to house guests. It is better to be warm and welcoming to guests when they’re around (makes them want to stay!) than ask them what happened when they’re leaving. Once customers have made up their mind to leave your product, they hardly stop to explain.

It's easier to tighten the embrace by building a product adoption model and taking them to higher levels of adoption step-by-step through a methodical journey. Increasing the exit barrier through such processes, listening to your customers, and making the product sticky is a more scalable and sustainable approach, than merely firefighting with the help of models.

In my experience, these proactive steps have served us much better than a model ever did. What has your experience been as you managed churn in your organization? I’d love to hear about it!