How you can use churn prediction to improve retention
Why are you losing customers? When are you losing customers?
When starting a business, the main goal is to bring in as many customers as possible, probably through extensive marketing and campaigning. But, at some point, the next biggest hurdle becomes keeping these customers so that your business doesn’t limitlessly expend what may be a limited budget on acquiring new customers.
Solely focusing on new customers is like filling water into a leaking bucket; your business won’t grow if you’re losing customers as much as you’re gaining them. To keep an existing customer, you have to recapture their attention before they leave– but how do you know when that’ll happen?
This is the perfect time to utilize churn prediction models to improve retention.
Firstly, let’s define what churn means. Churn is the measure of how many customers stop using a product over a specific period, and this measurement can be found in many ways: failure to renew a subscription, cancellation of a service, inactivity, etc. over the course of weeks, months, or even years. It’s the percentage of how many customers are lost.
A high churn rate means that many customers are leaving, and a low churn rate means that your business is retaining most of its customers. For a company to grow, the churn rate must be lower than the number of new customers gained (growth rate). Ultimately, the goal is to reach a churn rate as close to 0% as possible.
A churn prediction model is a machine learning model that tries to predict when and why your customers might leave based on what you already know about them. For example, Spotify could use your age and listening activity to predict how likely you are to churn next month.
Generally, this is how churn prediction models work:
The previous customer data would be some existing, historical data about your customers such as demographic information, their subscription/service type, prices, etc.
The machine learning model would identify connections using variables categorized as target and feature variables. A target variable is the expected effect that responds to other variables (ex: the cancellation of a Spotify Premium plan) and a feature variable is the expected cause that explains the result, such as the age or income of the customer. By organizing data as variables, the model is able to make predictions about the likelihood of new customers churning. Thus, it is called the churn prediction model.
The primary goal of churn prediction models is to identify when your customer will churn as early as possible so that you have more time to prevent the churn and retain your customer.
As the famous saying goes, you can’t make everyone happy. But, you just might be able to if you segment your customers and carefully strategize a customer retention plan for each segment. Because each person has their own set of interests and needs, one size does not fit all.
Tools like Cotera can help break down your customers into three different buckets: active, new, and inactive (churned), from which customers are further segmented into various categories based on their spending habits and activity.
Cotera actually has a case study discussing how their real-time customer segmentation model reduces churn rate that you can read more about here.
Once your model has forecasted which customers are most likely to churn, it’s up to you to prevent them from leaving.
There are a few customer retention strategies that companies use depending on their business model such as segmented campaigns, personalized retention offers, loyalty programs, surprises, and content recommendations– all of which will be further explained. Let’s continue using the world’s all-time favorite streaming service, Spotify, as an example.
Segmented campaigns are marketing tactics that target a specific group of customers. This could be special offers, product suggestions, or recommendations that aim to re-attract the segment at risk of churning. Spotify identifies its basic customer segments from the beginning, right when its customers discover Spotify for the first time.
At the very fundamental level, Spotify has two types of users: those on the free, ad-supported plan and those paying for premium, ad-free plans. To ensure that paying customers don’t leave– and that those on the free plan eventually convert to paying customers– Spotify offers numerous paid plans: Premium Individual, Premium Duo, Premium Family, and Premium Student.
These plans offer the same essential privileges of listening ad-free, just at different price points. For customers who may have less disposable income, such as students, Spotify offers Premium Student which is nearly half the cost of the individual plan. Spotify’s family plan allows up to 6 premium accounts under a single payment, and Spotify’s Duo encourages couples and friends to be avid users as well. This strategy works so well because even if one person under the plan has “churned,” the rest of the plan members are still active– thus the churn rate is not affected.
Spotify is also constantly releasing new products and services apart from music: audiobooks, podcasts, live events recommendations, and curated playlists to cater to various types of listeners and interests. If a customer gets bored of solely listening to music, they have other options such as podcasts. If a customer suddenly prefers listening to music live, they could still choose to keep their subscription to get first dibs on concert tickets from Spotify’s recommendations.
To minimize costs, companies could choose to offer promotions to certain individuals rather than their entire customer base. Loyal customers would continue to support the company regardless of promotions, so there is little benefit in offering them promotions as well.
Have you ever gotten an email from Spotify offering 3 months of free premium and then thought, Sweet! But then during those three months, you forget to cancel your subscription just to realize you’ll never be able to go back to a listening experience interrupted by ads every 3 minutes. This is how Spotify acquires new, paying clients, while concurrently reducing their churn rate [in the sense that users on the free plan don’t leave because they get tired of the ads].
Spotify wouldn’t offer everyone this sweet deal. Customers already paying for the premium plan wouldn’t need this offer, and Spotify might end up losing money instead. Rather, they specifically target users that they want to convert into paying customers for the long term.
Loyalty programs are incentives to attract and retain customers. Through rewards and discounts, customers are encouraged to return and become loyal. This strategy is also a great way for companies to identify which products and offers are attracting the most customers.
While Spotify doesn’t have a distinct loyalty program, hundreds of companies utilize this tactic to reduce their churn rate: think of places like Macy’s, Starbucks, Dunkin’, etc. These programs encourage their customers to return and spend more money to collect points to redeem.
Let’s take a closer look at Dunkin’s loyalty program.
Dunkin’ Rewards uses star rewards points that convert to a dollar amount. Customers are encouraged to spend more money to accumulate stars, which are then translated as dollar values to collect rewards in the form of free donuts, espresso shots, etc. The conversion rate is not exactly too generous, but it still works effectively to build customer loyalty.
This strategy is fairly simple: an occasional gift, coupon, or just a kind message that aims to surprise and delight customers.
Spotify regularly sends personalized concert recommendations, thank-you notes from commonly played artists, limited edition merch, and early pre-sale access to concert tickets– all of which are only available through Spotify.
Receiving these messages helps customers feel seen and heard; Spotify just gets it, you know? Nothing feels better than receiving a personalized thank-you letter from your #1 artist with an early invite to their upcoming concert.
Most companies have already incorporated recommendations to various corners of their platform, especially on e-commerce websites and social media apps.
Aside from generating playlists for segmented customers based on music genre, mood, shared artists, etc, Spotify heavily uses personalization to curate playlists that only you would like based on your listening patterns. With daily and weekly playlist recommendations as well as the famous, annual Spotify Wrapped, the app is quite irresistible and hard to abandon for current users.
There are numerous causes of churn such as financial concerns, competition, product defects, etc. Although it may be impossible to have a churn rate of 0%, it is important to persistently work towards a lower churn rate in order to grow as a company.
To make this possible, you need to track your results.
How do you know if your churn prediction model is accurate, and that your retention strategies are effective? How many customers are you actually saving? Are you wasting efforts on groups that would’ve purchased from your business regardless?
It’s good practice to do holdout testing to track your customer retention. Holdout testing provides valuable information about how effective your marketing campaign or retention strategy is through comparison. As your business experiments with various approaches to reduce churn, holdout testing ensures you don’t put all of your eggs in one basket.
This is how it goes:
To see if your holdout testing to improve customer retention is effective, use common metrics applicable to both your holdout and test group. Depending on your business model, this could be the customer churn rate, retention rate, repeat purchase rate, revenue per customer, etc.
Here are some useful formulas:
If there are no significant differences, consider experimenting with alternative marketing and retention strategies. Or, you might have to re-evaluate how your customers have been segmented. To learn more, you can read how Cotera utilizes AI models to improve customer retention here.