Authors: Ivaylo Bahtchevanov and Maryann Kongovi

Why You Can’t Ignore Churn

Customers are watching their pennies now, just like the rest of us. This means at a time that companies are doing everything they can to grow revenue, they need to be on high alert to anything that could push their customers away. Too many notifications? Unsubscribes will jump. Not enough? Usage and engagement could flatline.

In this day and age, customers expect a highly tailored experience, so rarely does a one-size-fits-all approach to engaging with them work. What’s the ‘just right’ strategy for interacting with your users? This is where machine learning can shine. For example, by employing AI-driven models tuned for notification strategies, one popular use case, companies can deploy the optimal approach at a customer level. This means notifications that engage customers and grow revenue, as opposed to triggering unsubscribes.

There are different types of churn worth paying attention to. Subscription churn happens when customers explicitly unsubscribe from your service. Usage churn occurs when customers quietly stop using your service or no longer perform transactions on your platform. Revenue churn occurs when you start making less revenue off of your existing users.

All of these examples tend to creep up on you without warning, and are inevitably a big hit to your business in the long term. Once a customer has left you, you may never get them back. Bain & Co estimates the cost of acquiring a new customer is typically five times more expensive than the cost of retaining an existing customer. This means that improving customer retention by 5% can lead to increased profitability of 25-95%. Similarly, the success rate of selling to a customer you already have is 60-70%, but this drops down to 5-20% when selling to a new customer.

Customers will rarely raise their hand to say they’re unhappy and thinking of leaving your platform, and only a small fraction will ever report when they have a bad experience with your product or service.

Keep Customers Coming Back with AI-driven Churn Prediction

Predictive analytics can help you identify which customers are likely to stop using your service altogether, and then deliver a highly personalized, optimized approach for re-engaging them. This is done through the use of f churn and retention models.

Churn prediction models identify who is likely to stop using your product within a specified timeframe, say within the next week or month, or the next quarter. You can predict when a user is likely to unsubscribe or uninstall the app, or changes in activity - sudden decline or drop in views, clicks, or purchases.

Once you identify the at-risk users, the next step is to understand why each user will churn. This can be crucial feedback for product teams to improve the customer experience. In some cases, the predicted churn reason is rooted in missing features or functionality, or there is some misalignment between the expected user behavior and the actual user behavior. Once you uncover the root cause, you can prioritize building the right set of features for your most impactful users.

Churn models can also segment users likely to churn by other factors such as lifetime value, which allows you to develop personalized retention strategies and prioritize your most profitable or highest spending users. The next step is to predict what each user’s brand/category/product affinities are. This will enable you to have an attractive targeting outreach. For example, an ecommerce platform would reach out with a series of recommended products or product pages the user is likely to be interested in. A streaming service would share a curated content list the user would like based on their viewing history. Then, you would predict which subsequent actions each user segment is likely to respond positively to (for example, does the user have a high response rate from mobile notifications or email), allowing you to quickly and cost-efficiently turn the high-churn risk users into highly engaged users.

Why use Kumo for Churn and Retention

Kumo employs the latest innovation in the world of machine learning - Graph Neural Networks. By treating your business data as a graph, Kumo’s models learn the contextual nuances behind the behavior of each of your users. You can read more about why GNNs are the latest innovative approach for dealing with your business data in our blog here. When looking to forecast and analyze churn,

Kumo builds your graph once, and allows you to run any number of predictions, for any number of time horizons, on your users and expand on your growth use cases quickly, with no additional overhead. This means you can quickly drill several levels down into your users’ behavior.

The ability to run many predictions quickly and concurrently leads to another major advantage of Kumo - the ability to perform what-if analysis. Think of this as simulating future user behavior to quickly test many hypotheses - you run through many actions and predict many future outcomes of how users will respond. You can predict any aspect of future customer activity in one single sitting. For example, Kumo allows you to predict across all of your high-risk users how they would react if you send them a coupon this week. Now do the same with a set of notifications - you can optimize the right outreach (mobile, desktop, email, etc) method and the right number of notifications (Read more about Kumo’s capabilities on churn and retention). You’d further personalize your outreach by sending the user curated products or content that they have a high affinity for.

With Kumo, you no longer have to guess what the outcome of a business strategy will be, nor do you have to perform live experiments that put your customers at risk just to understand them better.

Request a demo today!

Cookies settings