We have worked with hundreds of leading retail brands that all struggle with the same issue. They invest a lot of marketing resources in acquiring new customers. But they struggle to find effective ways to get those customers to stick around and make repeat purchases. Often these brands have programs in place for churn prevention, but these programs are typically one-size-fits-all, based on fixed business rules that don’t take into account the needs and behaviors of each individual customer.

In this guide, we are going to show you a better way to improve retention and reduce customer churn. It is based on the use of predictive analytics and machine learning to identify each customer at risk of churn, and continuous market testing to optimize retention while minimizing discounts. We’ll break it down into three phases—crawl, walk, and run—that start simple and become more complex as value is delivered.

But before get into operational details, let’s take a minute to understand why rules-based churn prevention programs aren’t good enough, and learn how predictive analytics revolutionizes customer retention efforts.

Business rule-based churn prevention programs aren’t good enough

Retailers using a one-size-fits-all approach might rely on “no purchase for 6 months” as their definition of customer churn. However, imagine a high-value customer who shops every week—by waiting 6 months to contact her, you’ve almost certainly missed the window to win her back. On the flip side, you’ve got a more typical holiday shopper who only buys once a year. If you reach out to her with a discount at 6 months to win her back, you’re throwing away discount margin—she wasn’t really at risk of churning in the first place! Either way, business rules that rely on the average of customer behavior leave money on the table; no two customers are alike, and the key to successful churn prevention is recognizing what makes each shopper unique.

Predictive Analytics Changes the Game

Predictive analytics makes it easy to identify customers at risk of churn. And the integration of predictive analytics with marketing campaign execution tools makes it easy to win them back at the right time with the right message. Here is how it works:

1. Identifying the right moment in each customer’s lifecycle. Predictive churn detection algorithms identify which customers look like they are at risk of fading away. The analogy we often use is “pattern matching”: machine learning models look for common threads across all different types of patterns of customer behavior. For example, what are some of the attributes associated with churning high-value customers? What do they tend to buy, where do they tend to live, and what does it look like when they show signs of fading away? By recognizing these patterns, advanced customer analytics tools are able to accurately predict what any customer is likely to do – by identifying good “lookalikes” within the customer database.

2. Turnkey campaign setup and Email Service Provider (ESP) integration. After identifying the lifecycle status of each and every customer, predictive analytics makes it easy to pinpoint key “thresholds” or moments in a customer’s lifecycle as he or she signs of fading away. Custora’s solution goes one step further and automate the delivers those segments directly to ESPs, complete with multivariate test setup.

3. Test-and-learn. Custora’s platform provides advanced reporting on incremental campaign results highlights what’s working and where there’s opportunity to keep improving the program.

4. Optimizing with advanced segmentation. Custora’s platform provides access to advanced behavioral segmentation (e.g., customer personas and predicted lifetime value tiers) to further tailor the content of churn prevention messages—driving additional relevance.

A Crawl, Walk, Run Operational Plan

As with many marketing programs that leverage segmentation, Custora advocates a “crawl -> walk -> run” approach that starts simple (to reduce operational complexity and resource needs) and progressively becomes more complex as value is proved.

Crawl

The starting point is to identify all potential levers that might be used for the churn prevention program. Here are some examples of commonly available levers:

Next, select 2-3 offers across a wide range. For example, if focusing on percent-off discount, a retailer might start with the following offer strategy:

One popular tactic to help anchor the range for the retailer is to ask the following thought experiment: “If you knew that a high-value customer was on the verge of churning, what discount would you feel comfortable giving to save the relationship?” Their response represents the high end of the range and can help find the right starting point.

Once the retailer has identified the range of offers or calls to action for the campaign, it’s time to hit the ground running with initial testing. The initial test begins by identifying all customers who fall within particular lifecycle status segments: all Cooling Down, At Risk, Highly At Risk, and Lost segments. Users within each of these segments are split evenly across a no message (holdout) group, and any calls to action that have been identified.

If you were working with us, the Custora team would set up the initial segments and test configurations within the software, and then publish those lists directly to your ESP.

Once the initial test goes live, it’s time to review the campaign results. Here’s a common potential outcome:

The goal here is to identify the optimal message/offer for each lifecycle stage. Then every week customers that newly cross into a given lifecycle status will get the optimal message automatically through email triggering.

Walk

In the “crawl” state described above, the goal is to identify the optimal communication strategy for each lifecycle status segment (One or More Purchases but Cooling Down, At Risk, Highly At Risk). The goal of the “walk” state is to get even more granular—to carve out different segments within each of these lifecycle stages to receive different offers.

A key enabling technology here is Custora’s ability to show segment-level campaign results after the campaign has been run – even if those segments weren’t identified upfront. For example, marketers can look at high-value customers’ responses to different messages or discounts, even if the high-value customer segment wasn’t specifically carved out at the time that the campaign was initiated.

For example, let’s say that we had the following lifecycle status-level data:

But when we looked at the response rate for different value-based customer segments, we observed the following:

Wow! This is telling us that while our low discount communication strategy is successful overall, it’s not really resonating with our highest-value customers. Perhaps we want to begin testing an alternate value proposition or call to action for those segments.

Previously, we were running a single test (low discount vs. control group) for the entire “One or More Purchases but Cooling Down” segment. After seeing the data referenced above, here is the testing strategy that a marketing team might try:

Once the relevant segments are carved out within each lifecycle status bucket—and the communication strategies configured—the weekly automation cadence will resume.

Run

In addition to offer or call-to-action optimization by segment, many retailers see results with creative optimization. This involves curating different brand experiences for different types of customers depending on what’s most likely to resonate with them: for example, showing sports imagery to “athletics” customers and runway imagery to “fashionista” shoppers.

Using cluster analysis is a great way to personalize brand experiences for different customers. Through advanced statistical techniques, Custora can quickly identify the customer segment cluster that each individual belongs to. Marketers then use Custora’s platform to examine these segment clusters to find unique insights about their purchasing patterns and demographics. We call these segments “data-driven personas” and they help marketers build creative that speaks to groups of customers with similar purchasing interests.

Segmenting by personas represents the “run” stage because it involves the creation of additional email content—in some cases, the creation of a new email for each individual persona. For example, a retailer with five robust personas (family shoppers, athletes, fashionistas, special occasion shoppers, and low-cost bargain hunters) might want to develop five versions of their email at each stage in the lifecycle. This introduces additional overhead to build and maintain, but can often lead to meaningful improvements in churn prevention performance.

Segmenting on both personas (for creative optimization) and value segments (for offer/call to action optimization) represents the “run” stage. At that point, Custora and the retailer should continue to refresh email creative periodically (e.g., quarterly) and look for additional optimization opportunities—but it’s time to transition the program into always-on, maintenance mode and look to the next set of incremental opportunities.

Final Thoughts

Customer acquisition and retention go hand in hand—in order to successfully transform customer economics, retailers need to focus on cultivating long-term relationships with the customers they’ve fought so hard to acquire. A platform powered by predictive analytics, that allows marketers to identify the right moment in each customer’s lifecycle, integrates with marketing channel execution tools, provides a test-and-learn environment, and is optimized with advanced segmentation capabilities is key to improving customer retention.

About Custora

Custora is an advanced customer segmentation platform.

We help marketers improve the ROI of their email, display, direct mail, and Facebook campaigns. Our platform give marketers the power to create smart customer segments and activate them everywhere.

Custora aggregates customer data from disparate sources, leverages predictive analytics to surface the optimal segments, integrates with marketing channels, campaign tools, and personalization technologies for easy segment activation, and measures the impact of each campaign at the customer level.