It is hard to think of a time when retail customer behavior saw more upheaval than it has in recent months. Long-standing customer routines have been disrupted. Online sales have grown dramatically. Category and brand preferences have shifted. And it has all happened at an unprecedented pace. As a recent Fast Company article declared, “You Don’t Know Your Customer Anymore.” The old rules – and the old data models – no longer apply.

At first, the focus was on nothing but execution. Shelves had to be filled. Supply chains needed to be re-tooled. E-commerce capacity needed to be scaled up. Now, several months into the global COVID-19 pandemic, many are turning their attention to trying to use retail analytics to make some sense of what has happened with the customers.

This seems like a good time to revisit a topic we have discussed in the past: customer churn. Even during more stable times, all retailers are losing some customers and gaining others. But as with many things these days, the patterns are more accelerated and more extreme. As customer habits change around how they cook and feed their families, how they work, and how they shop, it seems that nearly every customer decision is up for re-evaluation.

Most retailers have processes in place to track and reach out to lapsed customers. But these systems almost always focus on identifying churn at the store level. In other words, customers are flagged only when they have stopped shopping altogether. But often these efforts turn out to be too little, too late. Worse, they miss the many opportunities to grow sales with customers who may still be shopping with you but have stopped buying a certain category.

This is where churn at the category level comes in. By tracking – and even predicting – churn at the category level, retailers can respond before it’s too late. As the chart below shows, looking only at a customer’s total spend can easily mask underlying problems. Keep reading for our tips on how to make the most of category-level churn.

Customer Churn

1. Win them back while they are still your customer

There are many reasons a customer might stop shopping at your store. They might have moved to another town. They might be dissatisfied with your service. Or a competitor may have wooed them away. But one thing is certain: it will be a lot harder for you to communicate with them and get their attention after they have left.

One of the great advantages of measuring churn at the category level is that, in most cases, the customers are still visiting your store. If a customer has stopped buying beauty products from you, but still visits regularly for core grocery items, you still have a fighting chance. By the time you identify that a customer has completely disappeared from your store for a few weeks, a basket coupon with a “Please come back” message might not be enough.

Our analysis of data from retailers around the world shows the opportunity is significant. Take a look at the example below from a leading retailer. At the level of the total store, 80% of customers have low churn risk and 20% have high risk. But if we look only at the customers who appear stable at the total store level, we find that there are multiple categories potentially at risk. Depending on the category, between 13% and 19% of the apparently stable customers are actually at risk of stopping purchases in the category.

Customer Churn

2. Predict churn before it happens

Looking at categories rather than the total store is a great start in your loyalty marketing efforts. Even better is to identify the early warning signs that let you know a customer is at risk of churning from a category. A cookie-cutter approach will not work. This requires a robust set of models that take into account a customer’s past patterns and the unique patterns associated with the category. A customer who has not purchased shampoo or laundry detergent for six weeks should not raise any flags – these items have long consumption cycles. A customer who used to buy fruit and vegetables every week and suddenly stops would be a cause for concern after only a few weeks.


3. Understand your channels

Prior to COVID-19, customer behavior was already shifting between offline and online channels. In recent months, these trends have been accelerated on a massive scale. A solution that doesn’t give you a total view of your customer across all channels is no longer an option.


4. Automate it

It might be possible to manage a list of lapsed customers at the store level with some manual processes. Or to have the analytics team develop a churn model for a particular category for a special project. But to identify – and act on – churn signals for all customers across all categories on a regular basis, the process will need to be automated. A system that leverages machine learning and adapts over time as your customers’ behavior changes will take your churn prevention efforts to the next level.


5. Think beyond categories

We have focused here on identifying and preventing churn at the category level. But the ability to measure churn on a subset of your store doesn’t stop there. You can also think about tracking and predicting churn from brands (especially interesting for suppliers), from channels, or from other levels of your product catalog (department or sub-category). Especially interesting is the possibility of capturing churn from what ciValue calls “virtual categories” – these are collections of products that sit in different parts of the store but share a characteristic, such as Organic or Gluten-Free products.


The possibilities are endless. But one thing is certain. Retailers that are paying close attention to customer churn at a granular level will have a much better idea of what is happening with their customers – and a much higher chance of keeping them around.