Customer Data & Analytics Blog

How to Predict and Prevent Your Customer Churn

Ed Wolf | 4 minute read


Several weeks ago, we wrote about customer retention vs acquisition and discussed some pros and cons of each approach to customer marketing. Regardless of which direction is chosen (though ideally the best marketers are doing both effectively), customer churn or attrition is an ever-present danger. 

In this article, we will review the importance of preventing churn, and discuss some methods by which today’s savvy marketers are using data and AI to effectively predict and then prevent churn, while increasing customer engagement, loyalty, and ultimately lifetime value. 

The Importance of Measuring Customer Churn

In our customer retention vs acquisition article, we looked at some statistics showing the value of intelligently marketing to your existing customer base: 

● First, studies show that acquiring new customers is 5-7x more expensive than keeping new ones.

● Other figures along these lines show that an average customer lost can cost companies hundreds or even thousands of dollars that they may never recoup.

● Finally, another study shows that a loyal customer can ultimately be worth up to 10x the value of their first purchase with a brand.

In short, existing customers represent a major goldmine and should be treated as such. 

Therefore, the ability to predict that a customer is at a high risk of churning, and intervene in time to prevent it, represents a huge additional potential revenue source for most companies. Besides the loss of revenue that results from a customer leaving the business, the expense of initially acquiring that customer may not have already been covered by the customer’s activity to date.  In other words, acquiring that customer may have been a losing investment. 

Using data to retain

Churn prevention strategy rests on two pillars—understanding which customers are likely to churn and when, and then realizing which marketing actions are most likely to prevent it. 

The smart use of customer data can help with both pillars. First, companies need to study their customers’ historical data including items purchased, amounts spent, discounts, time between purchases, product affinity, items bundled together, web visits, shopping cart behavior, returns, in-store activity, reviews, surveys, social media mentions, demographic information, and many other such attributes. 

By doing this, companies can conduct thorough analysis of their customers’ behavior, ultimately segmenting them into categories or micro-segments. This type of detailed customer analytics goes far beyond such general categories as “New”, “Active”, or “Churn”. Rather, a micro-segment is a small and precise, yet relevant, group of customers, such as “Men who are between 35-40, from New York City, who typically buy dress shirts and shoes, and purchase twice a year, in-store, spending $200 each time, and have been customers for two years or more”. Once businesses have segmented their customer base into these small groups, they can look at the data to see exactly what these customers did over time, and then using algorithms and predictive models, understand what they are likely to do in the future. 

What they will discover from such an analysis is that while there are certain groups whose behavior patterns indicate that they will go on to be lifetime customers or VIPs with high Lifetime value (LTV) other groups will be likely to churn, in one, two, six, or twelve months. Others will be predicted to purchase once and never return. Once these categories are identified, a retention strategy can be formed and implemented. 

One key item is that this is not a one-time project, and this is where the value of AI and machine learning comes in. For churn analysis to be effective, the customer data, the micro-segments, LTV calculations, and churn predictions must be constantly updated.  This cannot be done manually, even with armies of data scientists, as there is too much data coming in from too many places to effectively analyze in a timely manner. As a result, data scientists will typically perform this analysis on an ad hoc basis rather than in a continuous, near real-time fashion. 

Thus, advanced customer analytics tools are becoming increasingly popular among brands as they struggle to retain their customers. To do this properly, AI is needed to curate, cleanse, and enrich the constant flow of data from the various sources, and then continuously recalculate the analytics to ensure there is no lag time between analysis and action. 

The next step after understanding who is likely to churn, is deciding what actions will be taken to get the customers reengaged and loyal. This is another area where micro-segmentation is a crucial step in the process. Brands need to identify specific sub-segments within the overall churn category to effectively perform 1-1 customer retention. 

For example, a generic “we’ve missed you” message, might work for a certain category of churned customers, but a more targeted “free shipping on item X” might be appropriate for a different group. Again, this is where data comes into play as analysis of historical data and A/B testing of offers will lead the way. Do mobile buyers respond differently than in-store customers? Does 10% off work just as well as 15% off? Segmenting and analyzing customer behavior with ALL the available customer data is the only way to effectively determine which retention offer will work for which customer via which channel. 

To summarize, for most businesses it’s very difficult and costly to acquire new customers—once they have done the hard work of attracting paying clients, to not focus on retaining them in a personalized, intelligent way flies in the face of logic. As we have seen, a churn prevention strategy can be a crucial revenue driver for brands. Companies that use their customer data, analyze it a meaningful way, and take appropriate action are winning this battle. 

To learn how Zylotech can aid with churn prediction and prevention efforts, please visit

Topics: Marketing Technology Customer Analytics