The AI experience: Customer clustering

When communicating with your customers, it’s important to know who you are talking to. It’s when communication takes the form of a real conversation, that things just click. But adapting your marketing communications to each customer would be an arduous and impossible task. So, what can you do? This is where customer clustering is crucial!

Clustering brings rapid descriptive insight about your customer base. A wide variety of different methods helps tailor results for any customer grouping use case, whether it is sales, web analytics or strategic planning.

Google Analytics describes that you have 200,000 new visitors this month. BI-reports tell you the story of your 10,000 active customers and their purchases. But when presented with the question “What are your customers like?” The business manager is often unable to answer. And especially when the question is presented as “What types of customers do you have?” or “What are your most important customers like?”.¨

When analyzing the customer database, it is simply not realistic to inspect each customer separately. While rich data for each customer is beneficial, it is impossible to describe an analysis showing each of the 10,000 customers simultaneously, while being useful in the decision making at the strategic level. Similarly describing all the customers in the customer base as one group without a sufficient level of granularity provides little insight. The middle ground here is to describe the customers as customer clusters or micro-segments.

The basic approach in customer clustering is to split the customers into groups simply using existing customer features such as

  • Demographics
  • Product purchases
  • Geographics

This is often a good start, but further on it does not give sufficient insight for strategic business planning or operations. Combining the customer data into meaningful groups will give you effective and actionable customer clusters with which to leverage the natural groups of your customer base.

Different applications of customer clustering

There are different methods of customer clustering. We go into some of the applications in more detail below:

  • Motive based segmentation
  • Customer base clustering
  • Website user clustering

Motive based segmentation:

Performing market research for the customer base is a good way of knowing the differences between your customer base and the whole market. Revealing purchase motives and triggers for your customers is a surefire way to boost the effectiveness of your marketing and sales efforts. The downside of this is that both are using samples and extending the motives into your existing customer base is not straightforward. Modeling the segments of a customer database predicts the motives of every customer and instead of knowing the segments at a general level, they can be attached to specific customers to further describe customer behavior even more. The reasoning and purchase motives of your customers can then be tightly integrated into your business plan starting from strategic focus areas to customer communication. Examining the differences between your own customers and the total market reveals strategic areas of focus for product development, sales, and marketing. Under-represented segments in the customer base are often the first places to look for potential business growth. Using these newly found segments in customer communication further enhances the relevance of your messages towards the customer.

Customer base clustering:

Even though questionnaires provide insight into customer purchase motives, it is not often a viable solution. Sending questionnaires requires a significant customer contact base, can be slow to collect, and sometimes might not be too linked to actual sales decisions. If the questionnaires and motives seem dubious then customer behavior data is the key, and if you got the data, clustering based on the customer actions could be the way to go. There are multiple different options for choosing the method of customer clustering, but the common factor is to align the clustering method with the business goals. If the only goal is to define targeting groups with transactional history, then more straightforward purchase focused clustering methods such as RFM-clustering might bring the best value. If a deeper understanding of customer behavior through service usage, communication and the field of business is needed, a machine learning-focused approach is often the way to go. Bringing the actual customer actions within clustering is usually a happy medium between strategic planning and more tactical sales-focused prioritization. Utilizing these clusters in your business processes is often somewhat easier than full motive-based segmentations. Of course, there are combinations of these methods (such as combining questionnaires with behavior data) that aim to bring the best of both worlds.

Website user clustering:

What if you could tailor your website to every user individually? Offering content the specific user is most interested in amongst the thousands of articles on the site. An individually tailored site like this would improve the user experience within the website and most likely increase site usage time. Although modern AI solutions enable us to individually personalize site experience, a more straightforward way is to cluster site users based on the sections they visit within the site. Taking site usage, possible customer data and purchase history into account, a website can be first tailored for the user group and through individual features then to an individual level.

Wrap-up with few key takeaways

Whatever the clustering method of choice is, the actual solution is often dictated by the business’s needs. Forming actionable and descriptive customer groups helps to get a grasp of the WHOLE customer base instead of relying on biases, gut-feeling and anecdotes. If you still are undecided on which type of clustering is the way to go, here are some useful pointers:

  • Demographics, geographics and industries are a good start
  • Determine the major use cases and available data for the clustering
  • If the use cases are for a specific application, a behavior focused cluster works better
  • If it is more important to know the motives and attitudes behind customer actions, motive-based segmentation is often a superior alternative
  • There are a wide variety of methods available for several different use cases
Liked this post? Share it with your friends
Share on facebook
Share on twitter
Share on linkedin
Ilari Vähä-Pietilä
December 10, 2020