How Cluster Analysis Can Help Retailers Improve the Customer Experience | Zebra Blog

The Analytics Capability Retailers Don’t Know They’re Underusing

Artificial intelligence and machine learning technologies can teach you a lot about your customers, and help them learn more about your offerings – if applied correctly.

A store associate shows a customer something on a tablet
by Guy Yehiav
February 21, 2020

Retailers are entering a new era of artificial intelligence (AI) and machine learning.  

That was abundantly clear at Shoptalk 2019, an annual trade show where retail executives gather to network and discuss new industry trends, technology, and innovation.

A couple of years ago, retailers were just beginning to notice and look into these technologies as “the latest shiny objects.” They now firmly understand the value and need for AI and machine learning. However, they’re still trying to figure out how best to implement practical applications of these technologies to align with their objectives. Notably, there were some in-depth discussions about how to apply AI and machine learning to customer personalization.

Personalization Beyond Demographics

Personalization is a critical part of the retail customer experience. Your customers don’t want to feel like cogs in a machine or just another face in the crowd — they want to know, without a doubt, that you value them and their business. This is where customer cluster analysis, a key capability of machine learning and AI-powered solutions like prescriptive analytics, comes into play.

Essentially, cluster analysis allows you to break down your customers into groups, or clusters, based on similar behaviors and characteristics (not just hard attributes like gender or zip code). From there, you can use this information to personalize customers’ experiences based on the parameters of their clusters, such as their favorite product attributes, loyalty, average basket sizes and more.

Advanced cluster analysis is a remarkably useful capability of machine learning and AI, and yet it’s underused.

At ShopTalk, many retailers shared how they had access to machine learning cluster analysis but were using it to cluster their customers by their demographics – such as age, location or household size – and not by their shopping behaviors. This is a huge miss.

In my experience, clustering by demographic tells you little, if anything, about how to personalize the customer experience (or how to encourage customers to spend more).

Think about it: the fact that you live at zip code 78701 doesn’t mean that you behave, purchase and feel the same way as your neighbors. There are countless characteristics beyond demographics that influence customer shopping behaviors, each of which can tell you much more about what each customer expects from a retail experience.

For example, imagine you are running a promotion on children’s jeans. Your first inclination may be to send the offer to a cluster of customers with children. It’s a logical approach. However, there’s more to consider when determining the target audience. Not every customer with kids is going to need, want or even like jeans for their kids – at least not right now. What if half of those customers’ kids attend private schools that require khakis or skirts? What if a large portion of the kids are girls who prefer dresses?

Either way, numerous people in that cluster of “customers with children” will receive a promotion that doesn’t fulfill their needs. The lack of personalization, or targeting, in your promotions can start to dilute your brand in your customers’ eyes. Too many generalized promotions may start to feel like spam. So, when you do actually send a targeted offer, they may be more likely to overlook it.

If you really want to drive sales, it would make more sense to characterize customers by typical buying patterns and send the promotion to those who buy the most children’s jeans or used to buy but stopped for some reason. Customers would receive more relevant promotions, and their experience would feel personalized.

Marketing to Customers? Machine Learning Shines.

Using machine learning and AI-based prescriptive analytics tools to “learn” what might actually appeal to customers, and then applying that information via targeted outreach, could be a big win for any retailer. However, the return on investment (ROI) proves to be especially significant when trying to increase basket sizes or sales frequency with repeat customers.

Here’s a real-life example:

A large general retailer, a Zebra customer, had a popular loyalty program based on spend amount. But it lacked the technology to offer personalized promotions to each loyalty tier. Our customer’s marketing department could only use blanket marketing across the entire loyalty program, which meant it gave discounts to people who would have paid full price anyway. The retailer approached Zebra to understand how prescriptive analytics could be applied in this situation and ended up adopting our Marketing module to get better visibility and control over its promotions.

The module now monitors customers’ loyalty statuses and sends prescriptive actions to the marketing team whenever a group of consumers is about to qualify for a higher tier. The prescriptive actions direct Marketing to send the customers targeted promotions based on typical buying-pattern behaviors. These promotions entice customers to make additional purchases in the categories they like and need to in order to push them over the threshold to the next loyalty tier. Since adopting the marketing-specific prescriptive analytics module, the retailer’s top-tier loyalty members have increased significantly. Its overall sales have increased as well since upper-tier customers are more likely to both spend more money and promote the brand to others. It’s been a win-win – and just more proof of the power of machine learning and AI, when applied correctly.

Just one tip: before selecting a prescriptive analytics tool for your retail environment, confirm that you will be able to cluster your shoppers with a variety of behavioral parameters. For example, can you segment/cluster based on average basket size? Product-return preferences (i.e. via mail or in store)? Loyalty ranking? Typical purchase types? Frequency of visits? Can you also cluster by general demographic information, such as household size or zip code? These are very important questions to ask, as advanced segmentation capabilities can greatly boost the effectiveness of your marketing efforts and customer experience personalization.

###

Editor’s Note:

If your retail or CPG organization is interested in improving your customer experience while also strengthening profits and margins, reach out to our retail analytics team or leave us a note in the Comments section below. Our team would be happy to talk to you about the analytics tools that can help you achieve your most ambitious goals.

null
Guy Yehiav
Guy Yehiav is a 25-year veteran of the supply chain industry and the General Manager of Zebra Analytics.